ChatGPT gets its answers from two places: the knowledge baked into the model during training, and live web retrieval that runs through Bing's index when you ask something current. If you want your company in those answers, you need to know which sources feed both layers, because it's almost never your website.

That last part is what most B2B marketing teams miss. They optimize their own domain while the AI builds its answer from G2, Reddit, Wikipedia, and someone else's listicle.

Where Does ChatGPT Get Its Information?

Path one is parametric knowledge. During training, the model ingests web-scale crawl snapshots plus licensed content, and that material gets compressed into the model's weights. OpenAI has signed content deals with Reddit, Axel Springer, News Corp, and the Financial Times, among others. Wikipedia is heavily represented in training data across every major model. When ChatGPT answers without searching, it's drawing on this layer: a statistical memory of what those sources said, frozen at the training cutoff.

Path two is live retrieval. When ChatGPT search kicks in, it retrieves current pages through Bing's index, reads them, and synthesizes an answer with citations. This is why Bing indexing suddenly matters again after a decade of nobody caring. If Bing hasn't indexed your page, ChatGPT search can't retrieve it, full stop.

Other engines run their own stacks. Perplexity built its own retrieval infrastructure. Gemini pulls from Google's index. Same two-layer architecture, different plumbing.

The strategic implication: you're playing two games at once. The training-data game is slow. It rewards presence in sources that get crawled and licensed, and it compounds over years. The retrieval game is faster. It rewards pages that are indexed, structured, and answer-shaped today. Most companies are losing both without knowing either exists.

What Sources Do AI Engines Cite for Vendor Questions?

Here's where it gets uncomfortable for anyone whose AI strategy is "improve our website." For commercial queries, the "best X for Y" and "which vendor should I pick" questions your buyers actually ask, AI answers lean heavily on third-party sources. In the citation audits I run for B2B clients, the same source mix shows up over and over:

Review platforms. G2 and Capterra profiles, category pages, and comparison pages get cited constantly. These platforms have structured data, dense entity signals, and thousands of pages mapping vendors to categories. To an AI engine, a G2 category page is a pre-built answer.

Comparison listicles. "Best CRM for mid-market SaaS" posts on domains that already rank. When a model needs to answer a best-of question, an existing best-of article is the lowest-friction source available. The listicle's author just became your gatekeeper.

Reddit threads. OpenAI licenses Reddit content directly, and retrieval systems surface Reddit for almost any "what do people actually use" question. Practitioner threads read as unfiltered peer opinion, which is exactly what a model wants when synthesizing a recommendation.

Wikipedia. For entity-level facts, what your company is, what category it's in, who founded it, Wikipedia remains the backbone. If your company has a page, models treat it as ground truth.

Industry publications. Trade press, analyst coverage, and established niche blogs carry authority in their domains. A mention in a respected logistics publication does more for a freight-tech vendor's AI visibility than ten posts on the vendor's own blog.

Notice what's barely on the list. Vendor websites get cited for facts about the vendor itself, pricing, features, docs. For the question that decides deals, "which one should I buy," your own site is a minority source.

Why Do AI Answers Favor Third-Party Sources?

Because third-party validation beats self-description, and the models are built to know the difference.

Every vendor's website says the vendor is the leader. That's not information; it's noise with a logo. A model trained to produce trustworthy recommendations learns to weight sources that evaluate vendors over sources that are vendors. A review platform aggregating 400 customer opinions, a subreddit arguing about real deployments, an editor who ranked twelve tools: these carry evidentiary weight your homepage can't.

This mirrors how a smart human buys. Nobody's final diligence step is reading the vendor's About page. They ask peers, check reviews, read comparisons. AI engines have operationalized that instinct at scale. I covered why this makes Google rankings a poor predictor of AI visibility in why you can rank on Google and still be invisible in ChatGPT; the short version is that the source graph AI trusts and the link graph Google ranks are different graphs.

How Do You Earn Presence in Each Source Type?

Each source type has a distinct playbook. Treat them as separate workstreams, not one "AI SEO" bucket.

Review platforms: build the profile, then drive volume. Claim and complete your G2 and Capterra profiles with the exact category language you want AI engines to associate with you. Then run a systematic review-generation motion, because review count and recency affect whether you appear on the category and comparison pages models cite. A profile with nine stale reviews signals a vendor nobody uses.

Listicles: get into the articles that already rank. Find the "best X" posts that show up for your category and pitch inclusion. Some accept updates on merit, some want a relationship, some are pay-to-play. Evaluate each on its terms. Earning a slot in an established roundup is usually faster than trying to outrank it, and once you're in, every AI engine citing that page inherits your presence.

Reddit: participate legitimately or stay out. Astroturfing gets detected, by moderators and increasingly by the models themselves, and it burns trust you can't rebuild. What works is founders and practitioners answering questions in their actual domain of expertise, disclosed and useful. One genuinely helpful comment thread where your product comes up naturally is worth more than fifty planted mentions, and it doesn't blow up in your face.

Wikipedia and industry press: this is digital PR. You can't write your own Wikipedia page into legitimacy; notability comes from independent coverage. So work the causality in the right order. Original research, data reports, and expert commentary earn trade-press coverage. Trade-press coverage builds the citation base that makes a Wikipedia presence defensible and makes industry publications reference you unprompted.

One caveat: don't abandon your own site. It still owns the vendor-fact layer, and structuring it for retrieval matters. That's a separate discipline I break down in how to get recommended by ChatGPT.

Do All AI Engines Cite the Same Sources?

No, and this trips up teams who spot-check ChatGPT and call it a day. ChatGPT retrieves through Bing, so Bing indexing and Bing-visible authority shape its citations. Perplexity's independent retrieval stack surfaces a different mix, typically heavier on recent web content. Gemini inherits Google's index and its quality signals, so it correlates more with traditional Google visibility than the others do.

A company can be well-cited in Perplexity and absent from ChatGPT, or vice versa. Check all the engines your buyers use, and check on a recurring basis rather than once, because answers drift as indexes and models update. Doing this manually gets old fast; I've reviewed the monitoring platforms that track citations across engines in my rundown of GEO and AEO tools.

What This Means for Your GTM Strategy

Stop thinking of AI visibility as a content problem on your domain. It's citation-graph work: systematically building presence in the sources models already trust, so that when an engine assembles an answer about your category, the evidence for including you is everywhere it looks.

In Coherence Model terms, this is Mass. Every review, every listicle placement, every legitimate community mention, every piece of earned coverage adds weight to your entity in the exact places AI systems measure it. And unlike paid channels, it compounds. Sources cite sources; a trade-press mention feeds a Wikipedia citation feeds a training corpus. The vendors getting recommended by ChatGPT today started accumulating that mass two years ago.

If you don't know where you stand across these source types, that's the first diagnostic to run, and it's exactly what a structured GEO engagement starts with. Map your presence in the citation graph, find the gaps against the competitors AI already names, and work the source types in order of leverage.

Frequently Asked Questions

Where does ChatGPT get its information?

From two layers. First, training data: web-scale crawl snapshots plus licensed content from partners including Reddit, Axel Springer, News Corp, and the Financial Times, with Wikipedia heavily represented. Second, live retrieval: when ChatGPT search runs, it pulls current pages through Bing's index and cites them. Older or general-knowledge answers come from the training layer; current-events and product-comparison answers usually involve retrieval.

What sources does ChatGPT cite for "best software" questions?

Predominantly third-party sources: review platforms like G2 and Capterra, comparison listicles on established domains, Reddit threads, Wikipedia, and industry publications. Vendor websites get cited for facts about the vendor itself but rarely drive the recommendation. In my own citation testing across B2B categories, third-party sources consistently outweigh vendor-owned pages for commercial queries.

Why does ChatGPT mention my competitor but not my company?

The model has third-party evidence for your competitor and not for you. If they appear in review-site categories, ranked listicles, community threads, and press coverage while you don't, the engine has reason to include them and none to include you. It's a citation-graph gap, not a product-quality judgment, and closing it means earning presence in those same sources.

Does being indexed by Bing matter for ChatGPT visibility?

Yes. ChatGPT search retrieves through Bing's index, so a page Bing hasn't indexed can't be retrieved or cited in a search-augmented answer. Verify your site in Bing Webmaster Tools and confirm your key pages are indexed. It won't fix a weak citation graph, but it's a hard prerequisite for the retrieval layer.

Do Perplexity and Gemini use the same sources as ChatGPT?

No. Perplexity runs its own retrieval stack and tends to surface a different, often more recent source mix. Gemini pulls from Google's index, so it correlates more closely with traditional Google visibility. The same query can produce different vendor lists across engines, which is why monitoring only ChatGPT gives you an incomplete read.

Originally published at strategnik.com/thinking/where-chatgpt-gets-its-answers.

Your Google ranking measures how much Google trusts your pages. A ChatGPT recommendation measures how much the model trusts your company as an entity, and the two systems share almost no scoring criteria. That's why you can sit on page one for your category keyword and never once appear when a buyer asks ChatGPT who they should evaluate.

This post is the diagnosis. If you want the fix playbook, that's a separate piece: how to get recommended by ChatGPT. Here I want to walk through the mechanism, because until you understand why the systems diverge, every fix you attempt will be SEO tactics wearing a new hat.

We Rank Well on Google but Are Invisible in ChatGPT. Why?

Because you earned the wrong kind of authority for this channel. A page-one Google ranking is the output of a system that evaluates pages: links pointing at a URL, relevance signals on that URL, engagement with that URL. You built URL equity, and you built a lot of it.

An LLM answering "what's the best freight audit platform" isn't ranking URLs. It's assembling a shortlist of companies from what it learned in training and what it retrieves at answer time, then synthesizing a recommendation. The unit of competition is the entity, not the page. If the model doesn't have a clear, well-corroborated concept of your company, your beautifully optimized pages are inventory in a warehouse the buyer never enters.

This is the single most common disconnect I see in AI visibility work right now. Marketing teams read their Search Console data, see healthy rankings, and assume the AI channel is covered. Then they ask ChatGPT the exact question their pipeline depends on and watch it name three competitors. The ranking was real. It just measured something else.

How Does ChatGPT Actually Answer "Who Should We Use" Questions?

Two mechanisms, and your SEO program barely touches either one.

The first is parametric knowledge: what the model absorbed during training. When someone asks a recommendation question with browsing off, the answer comes entirely from this baked-in memory. If your company appeared consistently across the training corpus in comparison articles, community threads, industry coverage, then you exist in that memory. If your presence was mostly your own domain, you're thin. One well-ranked website is a rounding error inside a corpus that size.

The second is retrieval. ChatGPT's search mode pulls live results, and it does so through Bing's index, not Google's. Right away, some of your Google-specific equity doesn't carry. But the bigger issue is what happens after retrieval: the model reads a small handful of sources and synthesizes one answer from them. There is no page two. There isn't even a page one in any meaningful sense. There's a source set of maybe a half-dozen documents, and everything else is invisible.

And watch which sources make that cut. In my own testing across B2B categories, the citations skew heavily toward third parties: review platforms, comparison listicles, Reddit threads, Wikipedia, industry publications. Vendor sites appear far less often than their Google rankings would predict. It's worth noting OpenAI has content deals with Reddit and major publishers, which tells you where they believe answer-quality lives. I break down the full source landscape in where ChatGPT gets its answers.

Why Won't ChatGPT Cite My Own Website?

Because for recommendation questions, you are a structurally biased source, and the system treats you that way.

Think about what your category page actually says: you, describing why you're the best choice. Every vendor's page says the same thing about themselves. A synthesis engine trying to produce a trustworthy answer to "who should I use" gets nothing decision-useful from any of them, so it reaches for third-party validation instead. The G2 category. The analyst roundup. The Reddit thread where a practitioner with no stake in the outcome says "we switched to X and it actually worked."

This is the part that stings for teams with strong SEO programs. Your comparison page might outrank the listicle that ChatGPT cites. Doesn't matter. For "best X" and "who should I use" queries, the listicle is evidence and your page is a claim. The model weights evidence.

Your own site still matters, but its job changes. It's no longer the destination that wins the click. It's the canonical reference that confirms what the third-party sources say about you, and the place retrieval systems go to extract clean facts about what you do. Which most B2B sites are terrible at, and that's the next problem.

Why Does Keyword-Shaped Content Fail AI Retrieval?

Because it was engineered to win a ranking auction, not to be quoted.

A decade of SEO practice produced a very specific content shape: 2,000 words targeting a keyword, a slow wind-up intro, the actual answer diffused across twelve H2s so the page can rank for forty variants. That shape worked because Google rewarded comprehensiveness and dwell time. A human would scroll. A crawler would index the whole thing.

Retrieval systems don't consume pages that way. They extract passages, and they favor passages that directly and completely answer the question in compact form. A page where the answer is smeared across 2,000 words offers no single extractable passage. The system grabs a cleaner one from somewhere else, often from a site with a fraction of your domain authority, because that page said the thing plainly in two sentences.

So the content that made you rank is, passage by passage, some of the least retrievable material in your category. Nobody made an error here. You optimized correctly for the old scoring function. The scoring function changed.

Does the Model Even Know What Your Company Is?

Often, no. Not crisply. And this is the entity problem underneath everything else.

Run the audit yourself. Read how your company is described on your homepage, your LinkedIn page, your G2 profile, your Crunchbase entry, your founder's bio, and the last three press mentions. In most B2B companies I audit, those are six different descriptions. Different category language, different positioning, sometimes descriptions from two pivots ago still live on high-authority third-party pages.

An LLM builds its concept of your company by triangulating across all of those sources. When they agree, you get a sharp entity the model can confidently place in a category and recommend. When they conflict, you get a blur. And a model won't stake a recommendation on a blur; it names the competitor whose signals resolve cleanly instead. In Coherence Model terms, this is Mass failing to cohere: plenty of raw material about you exists in the corpus, but it's out of phase with itself, so it never compounds into an entity the model trusts.

Google was forgiving of this inconsistency because it ranked your pages on their own merits. LLMs are not, because the entity is the product.

What Transfers From SEO, and What Doesn't?

Your Google authority is not worthless. It's necessary and insufficient, and it's worth being precise about which is which.

What transfers: domain authority correlates with being treated as a credible source when your pages do get retrieved. Crawlability and site structure transfer, since retrieval still requires fetching your content. Structured data transfers. And the discipline of earning links transfers, because links are one of the corroboration signals that build entity confidence.

What doesn't: keyword rankings as a scoreboard. Page-level optimization as the unit of work. Content volume as a strategy. And critically, the assumption that your own domain is where the battle happens. In AI answers, most of the surface area that decides whether you get recommended lives on properties you don't control.

The context matters here. SparkToro's zero-click research found the majority of Google searches already end without a click to the open web, and Gartner projects a 25% decline in traditional search volume as buyers move research into AI tools. The channel you're winning is shrinking while the channel you're invisible in absorbs its traffic.

Here's my actual position. The teams treating this as "SEO with extra steps" are going to lose two years relearning it, because the mental model is wrong at the root: one system ranks pages, the other recommends entities, and tactics built for the first quietly fail in the second. The teams that win will be the ones that reallocate effort from their own domain to the third-party surfaces where models form their opinions, and who make their company legible as an entity before worrying about any individual page. If you want to know where you stand today, run your company through the AI visibility grader, then take the results to the fix playbook. Diagnosis first, though. It changes what you build.

Originally published at strategnik.com/thinking/rank-on-google-invisible-in-chatgpt.

llms.txt is a markdown file at your site root that gives AI systems a curated, token-efficient map of your most important content. Here's the honest status: some AI agents fetch it, no major engine has committed to using it, and most of what's written about it is either hype or dismissal. Neither camp is doing the analysis.

I'll do it here. What the file is, what the evidence actually shows, and where it should sit in your priority stack.

What Is llms.txt?

llms.txt was proposed by Jeremy Howard of Answer.AI in September 2024. The idea is simple. LLMs working with your site at inference time face two problems: context windows are finite, and your HTML is full of navigation chrome, scripts, and boilerplate that wastes tokens. So you publish a plain markdown file at /llms.txt that says, in machine-friendly terms: here's who we are, here's what matters on this site, here are direct links to the pages worth reading.

A typical file has a title, a short summary of the site or company, and organized link lists with one-line descriptions. Some sites also publish llms-full.txt, which inlines the full content of key pages into one large document an agent can ingest in a single fetch.

That's the whole spec. No protocol handshake, no registration, no tooling required. You could write one in a text editor before lunch.

How Is llms.txt Different From robots.txt and Sitemaps?

They solve different problems, and conflating them is the most common mistake in coverage of this topic.

robots.txt is access control. It tells crawlers what they may and may not fetch. It's a fence.

An XML sitemap is inventory. It lists every URL you want indexed, with no judgment about which ones matter. It's a phone book.

llms.txt is curation. It tells a model which handful of pages best represent your company and what each one contains, formatted for a machine that pays by the token. It's a briefing document. robots.txt manages crawl-time access; llms.txt is aimed at inference-time consumption, when a model or agent is actively trying to answer a question using your site.

None of the three replaces the others. If llms.txt matters at all, it matters as a third layer, not a substitute.

Does llms.txt Actually Work? What the Evidence Says

This is where I have to be straight with you, because almost nobody else is.

The adoption story on the publishing side is real. Developer-tools companies moved first. Anthropic publishes llms.txt files for its documentation. Mintlify rolled the format out across the docs sites it hosts, which put it on thousands of sites at once. If you spend time in developer documentation, llms.txt is close to table stakes.

The adoption story on the consumption side is much thinner. There is no public confirmation that OpenAI, Anthropic, Google, or Perplexity systematically fetch and use llms.txt for retrieval in their consumer products. Server logs at various companies show some AI-agent fetches of these files, which tells you the file gets read sometimes, by something. It does not tell you the file influences what ChatGPT says about you when a buyer asks who solves your problem.

Google's John Mueller has publicly compared llms.txt to the old keywords meta tag: a signal sites want to send that engines may simply ignore. That comparison stings because it's apt. The keywords tag failed precisely because it let sites describe themselves, unverified, and self-description is cheap. Any format that says "trust my summary of my own content" carries the same structural weakness.

So the honest scorecard reads like this. Publishing adoption: real and growing in developer ecosystems. Consumer-engine adoption: unconfirmed. Measured citation impact: nobody has credibly demonstrated one. If a vendor tells you llms.txt moved their AI visibility numbers, ask for the mechanism. They won't have one.

Why Ship llms.txt Anyway?

Given all that, my position might surprise you: ship it. Three reasons.

The cost is near zero. This is an afternoon of work, most of which is thinking, not typing. There's no build pipeline to maintain and no downside risk. A static file that engines ignore harms nothing. Very few bets in marketing offer optionality this cheap.

The exercise itself has value. Writing a good llms.txt forces you to answer questions most B2B teams have never answered crisply. What is this company, in one machine-readable paragraph? Which ten pages actually define us? What does each one say? That's entity-definition work, and it pays off everywhere AI systems evaluate you, whether or not this particular file gets read. Half the companies I audit couldn't produce a clean answer to "what are your ten most important pages" today.

If adoption lands, you're early. The file format is trivial to consume, agentic browsing is growing, and the problem llms.txt addresses is real. Standards sometimes go from ignored to assumed quickly. If that happens, you spent an afternoon. If it doesn't, you spent an afternoon.

Strategnik practices this. There's a live file at strategnik.com/llms.txt: entity description, the core framework, the service pages that matter. It took less time to write than this post.

Why You Shouldn't Rely on llms.txt

Here's the failure mode I actually worry about, because I've watched it happen with every shiny AEO tactic: a team ships llms.txt, checks the "AI optimization" box, and stops.

llms.txt does nothing for the actual drivers of AI visibility. When an engine decides whether to cite you, it's weighing entity authority: whether you exist as a recognized, consistent entity across the sources it trusts. It's weighing citations, whether third parties reference you, because AI engines run on consensus, not self-description. And it's weighing whether your content is answer-shaped, leading with direct extractable answers instead of burying them under throat-clearing intros.

A curated site map influences none of that. You can have a perfect llms.txt and be completely invisible in ChatGPT, because the model has no independent evidence you belong in the answer. The work that builds that evidence is slower and harder, which is exactly why teams reach for the file instead. I keep the full list of what actually moves citations in the GEO checklist; llms.txt is one line item near the bottom, and it's near the bottom on purpose.

If you want a read on where you currently stand before deciding where to invest, run your site through the AI Visibility Grader. It measures the signals engines demonstrably use, not the ones we hope they will.

So Should You Ship llms.txt? The Verdict

Yes. This afternoon, ideally. Write the entity summary, curate the links, publish the file, and treat the thinking it forced as the real deliverable.

Then be equally clear about what you just did: you sent a signal that no engine has promised to receive. The visibility your buyers actually experience, whether ChatGPT names you when they ask who solves their problem, is built from entity authority, third-party citations, and answer-first content. That's the actual LLM SEO discipline, and it doesn't fit in an afternoon.

Ship the file. Spend your real effort on the work that moves citations. Any vendor selling you the reverse priority is selling the cheap half of the job.

Originally published at strategnik.com/thinking/llms-txt-does-it-work.

To get recommended by ChatGPT, you need to become a recognizable entity with consistent brand signals, machine-readable structure, and presence on the third-party sources AI systems actually cite. That work breaks into six steps, and the order matters: fix your entity identity first, then your content structure, then go earn citations off your own site.

If you want the full diagnosis of why you're invisible in the first place, read why your B2B content doesn't show up in ChatGPT. This post assumes you already know you have the problem. It's the playbook for fixing it, sequenced so a marketing leader can hand it to a team on Monday.

One framing note before the steps. SparkToro's research found 69% of Google searches already end without a click, and Gartner projects traditional search volume drops 25% by 2026. The buyers you're losing aren't going to a competitor's website. They're asking a chat window, and the chat window is answering.

Step 1: Run an Entity Consistency Audit

Before you write anything new, make every description of your company say the same thing everywhere. AI systems triangulate identity across your website, LinkedIn, G2, Crunchbase, app marketplaces, and press mentions. When those sources disagree about what you are, the model hedges, and hedging means you get left out of the answer.

I've audited dozens of B2B websites, and this is the most common failure I find. The homepage says "revenue intelligence platform," LinkedIn says "AI-powered sales insights," G2 has you filed under a category you abandoned two years ago, and the founder's bio still describes the 2021 pivot. A human squints past that. A retrieval system reads it as four weak entities instead of one strong one.

The audit itself is a spreadsheet, not a project. List every property where your company is described. Record the exact category language, the one-sentence description, and the named competitors on each. Then rewrite all of them to a single canonical description: same category noun, same problem statement, same positioning language. This takes a week and costs nothing but attention.

Step 2: Ship Organization and FAQ Schema

Structured data is how you tell machines what your content means instead of hoping they infer it. Two schema types do most of the work for AI visibility: Organization schema on your homepage, and FAQ schema on any page that answers buyer questions.

Organization schema should carry your canonical name, logo, description (the same one from Step 1), founding date, and sameAs links to every profile you audited. That sameAs array is doing real work: it explicitly connects your scattered web presence into one entity, which is exactly the disambiguation problem AI systems struggle with.

FAQ schema goes on product pages, pricing pages, and posts built around questions. Write the question the way a buyer would actually type it, and make the answer complete in two or three sentences. Your developers can ship both schema types in a sprint. If your site runs on a modern framework, it's an afternoon.

Step 3: Restructure Content to Answer First

Every important page should answer its core question in the first two sentences, because retrieval systems extract answers from the top of the content, not the bottom. The classic B2B blog post opens with 400 words of throat-clearing about how the landscape is changing, then gets to the point somewhere around the second H2. That structure was tolerable for SEO. For AI retrieval it's fatal.

The fix is an editing pass, not a rewrite. Take your twenty highest-value pages. For each one, identify the question the page exists to answer, and move the direct answer into the opening. Convert H2s into actual questions where it reads naturally. Cut the wind-up paragraphs entirely.

Notice how this post opens. The first sentence is the answer. That's not a stylistic preference, it's the mechanism: when an AI system retrieves this page for someone asking how to get recommended by ChatGPT, the extractable answer is sitting right at the top.

Step 4: Earn Presence Where ChatGPT Actually Looks

Here's the part most teams miss: the majority of AI-visibility work happens off your own website. When ChatGPT recommends vendors, it's synthesizing from review sites, comparison listicles, Reddit threads, industry publications, and analyst roundups. I broke down the source landscape in detail in where ChatGPT gets its answers, and the short version is uncomfortable: your beautifully optimized product pages are a minority input.

So go earn presence in those sources deliberately. Claim and complete your G2 and Capterra profiles in the right category, and run a real review-generation program, because volume and recency both matter. Pitch your way into the "best X tools" roundups that already exist for your category; the authors update them, and they respond to well-argued inclusion requests. Show up in Reddit and community threads as a genuinely useful participant, not an astroturfer, because those threads get retrieved constantly.

This is slow, unglamorous work compared to publishing another blog post. It's also the highest-leverage step in this playbook. In Coherence Model terms this is Mass work: every third-party mention adds weight that pulls AI recommendations toward you long after the effort is spent.

Step 5: Publish Data Worth Citing

AI systems cite sources that other sources cite, so the fastest way into the citation graph is publishing something nobody else has. Original benchmark data from your product, an annual survey of your buyer persona, a named framework that gives people language for a problem they couldn't articulate. These are the assets that industry publications reference, and those references are what teach the model you're an authority.

You don't need a research department. A SaaS company sitting on usage data across a few hundred customers can produce a benchmark report that becomes the default citation for its category. One genuinely original data point beats fifty thought-leadership posts, because the data point gets repeated by other people, and repetition by others is the signal AI systems trust.

If you publish it, make it easy to cite. Clear methodology, quotable single-line findings, a stable URL. Give the stat a name if you can. Uncited research is just a PDF.

My Competitors Show Up in ChatGPT and We Don't. How Do I Fix That?

Reverse-engineer where the AI is learning about your competitor, then go earn presence in those exact sources. When ChatGPT names your competitor, ask it why. Prompt it directly: "What sources informed that recommendation?" Use the browsing-enabled modes and check the citations. You'll usually find the same handful of inputs: a G2 category they lead, two or three listicles that include them and not you, a comparison page they wrote, active Reddit threads where users mention them.

That output is your gap list, and it converts straight into a work plan. If they're in five roundups you're missing from, pitch those five authors. If their G2 category presence is stronger, that's your review program's target. If a Reddit thread from 2024 keeps surfacing, join the current conversations in that community. You're not trying to out-content them. You're closing a specific, enumerable presence gap.

I'd also test your own baseline before and after. The AI visibility grader is a free tool I built that grades exactly this: whether AI systems recognize your company, and how you compare on the signals that drive recommendations. Run it first so you know which steps in this playbook are your actual bottleneck.

How Do You Measure Progress?

Treat AI visibility like a channel with a scorecard, or the work will die after one quarter. The measurement is straightforward even though the tooling is young. Build a panel of 15 to 20 prompts your buyers would realistically ask: category recommendations, comparison questions, "best tool for" queries. Run them monthly across ChatGPT, Perplexity, and Google's AI Overviews. Track three things: are you mentioned, in what position, and with what description.

Add a second-order metric: is the description accurate? Getting mentioned with a stale or wrong positioning is its own problem, and it points you back to Step 1. Also watch your referral traffic from AI surfaces; it's small today but it converts well in my experience, because someone clicking through from an AI answer has already been pre-sold by the recommendation.

If this becomes a program rather than a project, that's the point where most teams need structure or outside help. It's the core of what I do on the GEO side of my practice, and the honest pitch is that the process above is not secret. The advantage goes to whoever executes it earliest and most consistently.

The window here is the interesting part. Entity authority compounds, and AI systems show a strong incumbency bias: once a vendor is established in the answer set, displacing it takes disproportionate effort. Right now most of your competitors are still writing keyword briefs for a search channel that's shrinking. Two years from now, the recommendation slots in your category will be largely settled, and the companies holding them will have done this work while it was still cheap. Do it while it's still cheap.

Originally published at strategnik.com/thinking/how-to-get-recommended-by-chatgpt.

Generative engine optimization is the work of making your company retrievable and citable by AI systems like ChatGPT, Perplexity, and Google AI Overviews. This checklist covers the 20 fixes that actually move citations, grouped into four tiers and ordered by impact, so you can stop guessing which GEO advice matters.

One honest framing before the list. Tiers 1 and 2 are days of work. A competent marketer with dev support can clear them in a sprint or two. Tier 3 is the months-long compounding work that determines whether AI engines ever name you, and it's where most companies quit. If you want the whole thing run with someone who does this every week instead of piecing it together from checklists, that's what generative engine optimization services exist for. Either way, the sequence below is the sequence.

If you want the strategic case for why this discipline exists at all, the Answer Engine Optimization page covers it. This post is the execution list.

Tier 1: Entity Foundation

AI engines retrieve entities, not pages. Before any content work pays off, the machines need to know what your company is, what category it belongs to, and that every source describing it agrees. This tier is mostly configuration. Do it first because everything downstream depends on it.

1. Add Organization schema to your homepage. Include your legal name, logo, description, founding date, and sameAs links to every official profile. This is the machine-readable declaration of what your company is, and it takes a developer under an hour.

2. Write one canonical entity description and deploy it everywhere. Your homepage, LinkedIn, G2, Crunchbase, and every directory listing should describe the company in the same terms. AI systems triangulate identity across sources; conflicting descriptions read as low confidence, and low confidence means you get skipped.

3. Claim your Google Knowledge Panel. If one exists, verify it and correct it. If one doesn't exist, treat that as a diagnostic: Google's Knowledge Graph doesn't recognize you as an entity yet, which predicts poorly for every other engine.

4. Standardize your name, address, and category language. Pick one company name format and one category phrase ("freight audit software," not five variations) and enforce them across every surface. Category consistency is how engines decide which question you're an answer to.

5. Publish an llms.txt file. This is an emerging convention: a markdown file at your root that summarizes your site for language models. Adoption by the major engines is unproven, and I've written about whether llms.txt actually works, but it costs an hour and the downside is zero. Do it last in this tier, not first.

Tier 2: Content Retrievability

Entity work tells engines what you are. This tier makes your content extractable when they go looking for answers. The pattern across all five items: AI retrieval pulls concise, direct passages, so structure every page as if a machine will quote 80 words of it. Because one will.

6. Answer the question in the first two sentences of every page. Not after the empathy paragraph, not after the "in a world where" setup. Lead with the answer, then support it. Pages that bury the answer get passed over for pages that don't.

7. Rewrite your H2s as questions buyers actually ask. "How much does freight audit software cost?" beats "Pricing Considerations" because retrieval systems match headings against real query language. Pull the phrasings from sales calls and search data, not from what sounds polished.

8. Add FAQ schema to your service pages. Question-and-answer markup is the most directly extractable structure you can give a retrieval system. Write real answers, 40 to 80 words each, in the same language buyers use.

9. Enforce one page per intent. If three pages half-answer the same question, engines can't tell which one is canonical and often cite none of them. Consolidate overlapping pages into a single authoritative one and redirect the rest.

10. Kill your thin pages. Tag archives, 200-word posts from 2021, orphaned landing pages. They dilute the site's signal-to-noise ratio for anything crawling it. Delete, consolidate, or noindex.

Tier 3: Citation Surface

Tiers 1 and 2 make you retrievable, but AI engines mostly decide who belongs in an answer by consensus across third-party sources: review platforms, comparison articles, community threads, cited research. I've mapped where ChatGPT actually gets its answers, and the short version is that your own website is a minority input. In the Coherence Model, this tier is Mass work: building weight that other sources reference, which compounds for months and can't be faked in a sprint. It is also the tier that separates companies that get cited from companies with clean schema and no mentions.

11. Get listed on the review platforms your category uses. G2, Capterra, TrustRadius, or whatever your buyers actually consult. Engines lean heavily on these for "best X" answers, and an absent profile is evidence you don't belong in the category.

12. Earn placement in ranked listicles. When someone asks an AI engine for the best tools in your space, it synthesizes from existing "best X" roundups. Identify the ten that dominate your category's results and do the unglamorous outreach to get included.

13. Publish original data worth citing. A benchmark report, a survey, an analysis of your own platform data. Other sites cite data, engines follow citations, and one genuinely useful dataset outperforms fifty opinion posts.

14. Name your frameworks. A named methodology is a retrievable entity; an unnamed process is a paragraph nobody can reference. If you have a real point of view on how your problem should be solved, give it a name and defend it consistently.

15. Build a legitimate presence where your category gets discussed. For most B2B categories that means Reddit, plus industry Slacks and forums. Community threads feed AI training and retrieval, but astroturfing gets detected and torched, so this means actual participation over months. Slow is the point.

Tier 4: Measure and Iterate

You can't manage what you never baseline, and AI visibility is noisy enough that unmeasured GEO work turns into faith-based marketing. None of this requires paid generative engine optimization tools to start. A spreadsheet and a repeatable prompt set will carry you further than a dashboard you check twice.

16. Baseline your AI visibility before touching anything else. Run a structured set of buyer-realistic prompts across ChatGPT, Perplexity, and AI Overviews, and record who gets named. The free AI visibility grader automates a first pass if you'd rather not build the prompt set yourself.

17. Track mention rate monthly. Same prompts, same engines, same cadence. The metric is simple: in what percentage of category-relevant answers does your company appear?

18. Monitor which sources power competitor mentions. When an engine cites a competitor, note the underlying source. That's your Tier 3 target list, written for you by the machine you're trying to influence.

19. Re-test after each fix batch. Ship Tier 1, measure. Ship Tier 2, measure. Attribution stays impossible if you change everything at once and check once a quarter.

20. Treat variance as noise and trends as signal. The same prompt returns different answers on different days; that's how these systems work. A single dropped mention means nothing. Three months of declining mention rate means everything.

Where to Start

Do Tier 1 this week. It's cheap, fast, and foundational, and there's no defensible reason a B2B company should still have inconsistent entity signals in 2026. Do Tier 2 over the next month as you touch each core page. Run item 16 before either, so you can prove the work did something.

But don't confuse finishing the fast tiers with doing GEO. Schema and answer-first structure are table stakes that your competitors will also complete, probably this year. Tier 3 is the moat, because citation surface takes months to build and months for anyone else to replicate, and the engines reward accumulated third-party evidence over on-site polish every time. The companies winning AI citations in your category two years from now started their Mass work early and kept going after the checklist stopped feeling productive. Start the compounding work now, measure it monthly, and let everyone else stop at schema.

Originally published at strategnik.com/thinking/generative-engine-optimization-checklist.

To check what ChatGPT says about your brand, open a temporary chat, ask the questions your buyers actually ask, run each one several times, and record whether you're named, where you land, and who beats you. Most companies skip the method, ask one casual question from a logged-in account, and walk away with an answer that's contaminated, unrepeatable, and wrong.

I run these audits for B2B SaaS companies, and the manual version takes about an hour. Here's exactly how to do it, and then the shortcut that does it for you.

Why Can't You Just Ask ChatGPT Yourself?

You can. But the way most people do it produces garbage data, for three reasons.

AI answers are non-deterministic. Ask ChatGPT "what are the best freight audit platforms" three times and you'll get three different lists. Sometimes the same five names in different orders. Sometimes a name appears in run one and vanishes in run three. A single prompt is one sample from a distribution, and one sample tells you almost nothing. If you run a question once and see your brand, you might be at 30% inclusion and celebrating.

Your account contaminates the results. ChatGPT has memory. It knows you work at your company, it's read your previous chats about your positioning, and it will happily name your brand because you keep bringing it up. That's not visibility. That's a mirror. The same applies to Gemini if you're logged into a Google account soaked in your own brand searches.

Logged-in context skews retrieval. Even without explicit memory, personalization signals nudge what gets surfaced. Your buyer doesn't have your history. Testing from your own context is like Googling your company name from your office network and concluding your SEO is great.

The fix is simple discipline. Use ChatGPT's temporary chat mode (or a logged-out incognito window), run every question at least three times, and test in the phrasing a buyer would use, not the phrasing you wish they'd use. Nobody asks ChatGPT "tell me about [your tagline]." They ask "what's the best X for Y."

What Questions Should You Actually Test?

Test the questions that sit in a buying process, plus one question that tests whether the AI knows what you are at all. When I build a question set, it looks like this:

Five buyer questions:

1. Category recommendation: "What are the best [your category] platforms?"

2. Qualified recommendation: "What's the best [your category] tool for [your ICP — e.g., mid-market shippers, B2B SaaS teams under 200 people]?"

3. Problem-first: "How do I solve [the problem you exist to solve]? What tools help?"

4. Competitor comparison: "How does [top competitor] compare to alternatives?"

5. Shortlist request: "I'm evaluating [your category] vendors. Who should be on my shortlist?"

One entity question:

1. "What is [your company name]?"

The entity question matters more than it looks. If ChatGPT can't accurately describe what your company does, the recommendation questions are unwinnable. You can't be recommended for a category the model doesn't know you belong to.

Run this set across at least three engines: ChatGPT, Perplexity, and Gemini. They retrieve from different sources with different weights. Perplexity leans hard on live web citations. Gemini pulls from Google's index and Knowledge Graph. ChatGPT blends training data with its own browsing. I regularly see companies visible in Perplexity and absent in ChatGPT, or vice versa, and the gap tells you which signals you're missing. One engine is not a read on "AI visibility." It's a read on one engine.

Three engines, six questions, three runs each. That's 54 data points instead of one.

What Should You Record for Each Answer?

Don't just eyeball the responses. Score them. For every run, capture five things in a spreadsheet:

1. Mentioned at all? Yes or no. This is your inclusion rate, and it's the headline number. Mentioned in 2 of 18 ChatGPT runs is a very different situation from 18 of 18.

2. Position in the list. First mention carries disproportionate weight. Buyers using AI for research rarely dig past the top two or three names, the same way they never clicked to page two of Google. Fifth on a list of seven is closer to invisible than it feels.

3. Described accurately? Read what the AI says about you. Right category? Right ICP? Current positioning, or the thing you pivoted away from two years ago? An inaccurate description in a recommendation is arguably worse than absence, because now the buyer is evaluating a version of you that doesn't exist.

4. Which competitors appear? List every vendor named, every time. After 54 runs you'll have a frequency table, and it's usually humbling. The competitor set AI engines carry is often not the competitor set on your battlecards.

5. What sources are cited? Perplexity and Gemini cite openly; ChatGPT cites when browsing. Note every source: G2 category pages, "best X" roundups, Reddit threads, analyst lists, competitor comparison pages. This is your treasure map. These are the specific places AI engines learn your category, written down for you.

How Do You Interpret the Results?

The pattern of failure tells you which problem you have. Three patterns cover most audits I run.

Absent everywhere, including the entity question. You have an entity problem. The AI doesn't know what you are, so it can't recommend you for anything. This usually traces to missing Organization schema, no Knowledge Panel, and inconsistent descriptions of your company across your site, LinkedIn, and review platforms. The full diagnosis is in why your B2B SaaS is invisible in ChatGPT, and it's fixable, but it's foundation work, not a content tweak.

Present but described wrong. You have a signal-consistency problem. The engines know you exist but they're triangulating from conflicting sources: an old About page, a stale Crunchbase entry, a G2 category you outgrew. The model runs something like a consensus algorithm on your identity, and the consensus is out of date. The work is consolidation. Make every surface say the same thing about what you are.

Competitors keep appearing, with citations. You have a citation-surface problem, and the good news is the audit just handed you the fix list. If ChatGPT recommends your competitor and cites a specific roundup, that roundup is where you need to be. If Perplexity keeps pulling from a G2 category page, your G2 presence is now a GTM priority. Go earn the exact sources your competitors are being cited from. This is the most actionable output of the whole exercise, and it's the core of how to get recommended by ChatGPT, the companion playbook to this one.

In Coherence Model terms, this audit is a Gravity measurement: it tells you whether your brand actually pulls demand in the channel where buyers now do their research, or whether you've been assuming it does.

Is There a Faster Way to Run This Audit?

Yes. I built the manual method above into a free tool: the Strategnik AI Visibility Grader.

It runs the same methodology automatically. It asks buyer questions and an entity question across multiple AI engines, crawls your site to check the technical signals that drive retrieval, and scores you on four dimensions of AI visibility. You get the inclusion data, the competitive picture, and the diagnosis without spending an hour in incognito windows with a spreadsheet.

I'd still encourage you to run a few manual prompts yourself at least once. Watching ChatGPT recommend your competitor to your exact ICP, in real time, creates urgency that no report can. But for a repeatable baseline you can re-run quarterly, use the grader. It's free, and it exists because I got tired of doing the spreadsheet version by hand.

Frequently Asked Questions

How do I check if ChatGPT recommends my brand?

Open a temporary chat (so memory and personalization don't contaminate results), ask 5 buyer-style questions plus "What is [your company]?", run each question three times, and record whether you're mentioned, your position, description accuracy, competitors named, and sources cited. Repeat in Perplexity and Gemini, since each engine retrieves from different sources. Or run the AI Visibility Grader, which automates this exact process.

Why is my brand invisible when people ask ChatGPT for recommendations in my industry?

Usually because the AI doesn't recognize your company as an entity in your category. If you're absent across all engines and ChatGPT can't answer "What is [your company]?", the cause is structural: missing schema, inconsistent brand descriptions across platforms, and no presence in the sources AI engines cite, like review-site categories and roundup articles. Google rankings don't transfer. A company can rank page one and never be named in an AI answer.

How often should I re-run an AI visibility audit?

Quarterly is a sensible cadence, plus after any major positioning change. AI engines update their retrieval sources continuously, so visibility shifts faster than traditional search rankings. Track your inclusion rate over time rather than reacting to any single run.

For the full framework behind this methodology, see the Answer Engine Optimization strategy page.

Originally published at strategnik.com/thinking/check-what-chatgpt-says-about-your-brand.

The best AI SEO tools in 2026 fall into four working categories: content optimization and drafting, keyword research and clustering, technical SEO assists, and the emerging class of AI agents that execute SEO tasks end to end. No single tool covers all four, and the vendors claiming otherwise are selling a content generator with extra dashboards.

There is also a boundary problem worth naming up front. If your actual goal is showing up in ChatGPT and Perplexity answers, that is GEO tooling, a different category with different buyers, covered separately in the best GEO and AEO tools. This post is about AI applied to traditional SEO: ranking pages in search results that still drive most B2B pipeline.

One idea threads everything below. AI SEO tools compress execution time, and that compression is real. But every one of them optimizes against the same models of the same SERPs. When the whole market runs the same optimization math, tool-generated output converges on sameness. Your differentiation has to come from positioning, proprietary data, and an actual point of view. No tool supplies those.

AI Content Optimization and Drafting Tools

This is the biggest and most crowded category, and the one where honest expectations matter most.

Surfer scores content against the pages currently ranking for a target term and guides structure, terms, and coverage. Clearscope does similar work with an emphasis on topical completeness and readability. Frase combines SERP-based briefs with drafting assistance. Jasper sits closer to the pure generation end: a writing platform with SEO workflows layered on, aimed at marketing teams producing at volume.

Here is the honest take. These tools optimize for a correlation model of what ranks. They look at pages that currently perform and infer the shared characteristics: terms used, questions answered, headings covered. That is genuinely useful for coverage. It catches the gaps a subject-matter expert leaves because the answer feels obvious to them.

It is dangerous as autopilot. A correlation model cannot tell you whether a page deserves to exist, whether your take differs from the ten pages it was scored against, or whether the resulting draft says anything a buyer would remember. Teams that let the score drive the writing produce content that reads like the average of the current SERP. The average of the SERP does not win the SERP, and it definitely does not win a buyer.

Use these tools to check coverage on drafts written by someone with an opinion. Do not use them to generate the opinion.

AI Keyword Research and Clustering Tools

Keyword research is where AI features have quietly become table stakes inside the established suites.

Semrush and Ahrefs have both folded AI into their core research workflows: intent classification, topic grouping, and assistance layered onto keyword databases that were already the reason to pay for them. If your team lives in one of these suites, the AI features are an upgrade to workflows you already run, not a new purchase decision.

Keyword Insights is the notable specialist. Its job is clustering: taking a large keyword export and grouping terms by whether they can be served by the same page, based on SERP overlap rather than string similarity. That distinction matters. Two keywords that look related can demand separate pages, and two that look unrelated can be the same intent. Clustering by live SERP behavior is the right method, and doing it manually at scale is miserable.

The limit of the category is that keyword tools describe demand that already exists and is already visible to every competitor running the same export. They will not find the problem your ICP describes in language no keyword database has indexed yet. That comes from sales calls, win-loss interviews, and support tickets. The best keyword strategy I have seen in B2B SaaS starts with what buyers actually say and uses the tools to size and validate it, not the reverse.

Can AI Actually Do Technical SEO?

Partially, and the partial version is worth having.

Screaming Frog remains the crawler of record, and its export-heavy workflow pairs unusually well with LLMs. Crawl exports fed into an LLM workflow can triage issues, draft redirect maps, and turn a wall of status codes into a prioritized fix list in a fraction of the old time. Screaming Frog has also added its own AI-adjacent features for working with page content at crawl time. Sitebulb approaches the same territory with built-in prioritization and explanations, which makes it friendlier for teams without a dedicated technical SEO.

The compression here is real and mostly safe, because technical SEO has right answers. A broken canonical is broken. The risk profile is different from content: an LLM misjudging issue priority wastes time, but it does not publish sameness to your domain.

What AI assists still cannot do is decide architecture. Whether your site should have twelve service pages or four, whether a content hub belongs at the subdomain or the subfolder, how LLM SEO considerations should shape your markup and structure so machines parse your business correctly. Those are judgment calls with tradeoffs, and the crawler just reports what it finds.

What About AI Agents That Do SEO End to End?

This is the emerging category, and the one changing fastest. Instead of a tool that assists a human task, agents chain the tasks: research a topic, draft the page, build internal links, monitor the result, propose the next move.

The capability is arriving faster than most teams' judgment about where to apply it. Agents are legitimately good at bounded, verifiable work: crawl triage, internal linking passes, metadata cleanup, refresh candidates. They are a liability when handed open-ended editorial authority, for exactly the convergence reason above, now at higher volume and lower supervision.

This category deserves its own treatment, and it gets one in AI agents for SEO, published alongside this post. The short version: agents amplify whatever system they are dropped into. A coherent strategy gets executed faster. An incoherent one produces mediocrity at scale.

If the Goal Is AI Visibility, You Are in the Wrong Aisle

A growing share of people searching for AI SEO tools actually want something else: visibility inside AI-generated answers. Citation monitoring in ChatGPT and Perplexity, entity and schema infrastructure, AI Overview tracking.

That is GEO and AEO tooling. Different tools, different vendors, different work. The categories above will not tell you whether ChatGPT cites you, and a citation-monitoring dashboard will not fix a page that cannot rank. If AI-answer visibility is the goal, start with the GEO and AEO tools roundup instead. Most B2B teams eventually need both, but buying one while needing the other is the most common tooling mistake I see right now.

How to Choose, By Team Situation

Solo marketer or small team producing content in-house. One content optimization tool plus the AI features inside a suite you already pay for. Skip agents until your strategy is stable enough to be worth accelerating.

Content team of three or more shipping weekly. Add clustering. At this volume, cannibalization and page-mapping mistakes cost more than the tool does, and SERP-based clustering prevents both.

Team with real technical debt. Crawler plus LLM triage before anything else. Fixing crawl and indexation problems usually outperforms another optimized post, and it is the cheapest win in this entire post.

Team already ranking, worried about AI search. Your bottleneck is not in this article. Go to the GEO roundup.

Team with budget and no strategy. Do not buy anything yet. Every tool here assumes someone has already decided what you should rank for and why a buyer should pick you. Tools execute decisions. They do not make them.

The Part No Tool Sells

Run the thought experiment: you and your three closest competitors all buy the same optimizer, the same clustering tool, the same agent framework. Everyone's execution speeds up. Nobody's position improves, because you are all optimizing against the same models of the same results with the same inputs. The tools cancel out.

What does not cancel out is the input only you have. Your positioning, your customer data, the argument your founder can make that competitors cannot. Tool-assisted content built on a distinct point of view compounds. Tool-generated content built on SERP averages fills a sitemap.

That is the difference between buying tools and having someone architect the system the tools plug into: what to rank for, what to say that nobody else can, and how content, structure, and authority reinforce each other. If the stack keeps growing while results stay flat, the gap is architecture, and that is the case for working with an AI SEO agency that builds the system rather than reselling the dashboards.

Frequently Asked Questions

What is the best AI tool for SEO? There is no single best AI SEO tool because the category spans four jobs. Surfer and Clearscope lead content optimization, Keyword Insights leads clustering, Screaming Frog with LLM workflows leads technical triage, and agents are too new to have a durable leader.

Do AI SEO tools actually work? Yes, for compressing execution: briefs, clustering, and crawl triage all get meaningfully faster. They do not work as a substitute for strategy, because every buyer of the same tool gets the same recommendations.

Will AI-generated content rank? It can, when it carries real expertise and a distinct angle. Pure tool output tends to converge on what already ranks, which makes it easy to publish and hard to prefer.

Originally published at strategnik.com/thinking/best-ai-seo-tools.

An AI SEO agent is software that plans and executes multi-step SEO work toward a goal: crawl the site, find the issues, draft the fixes, open a pull request. A tool assists with a task you drive; an agent runs the sequence itself and checks in with you at defined points.

That distinction matters more than any feature list, because the market is currently flooded with products calling themselves "SEO AI agents" that are really tools with a chat interface bolted on. Before you evaluate one, get clear on what you're actually buying and what you're actually delegating.

What Is an AI SEO Agent?

A tool waits for you. You run the crawl, read the report, decide what matters, make the change. The tool did one step; you did the workflow.

An agent owns the workflow. You give it a goal — "find and fix broken internal links across the blog" — and it plans the steps, executes them, and surfaces its work at checkpoints you define. On the technical side that looks like crawl → diagnose → draft fixes → open PRs for review. On the content side it looks like research → brief → draft → interlink → queue for human approval.

The checkpoints are the difference between an agent and a liability. Any vendor selling a fully autonomous loop with no review gate is selling you a way to ship mistakes at scale.

If you're still at the tool stage, that's fine and often correct. I've written separately about which AI SEO tools are actually worth using, and for plenty of teams a good tool plus a sharp operator beats a mediocre agent.

What Can AI Agents for SEO Actually Do?

The honest list is shorter than the marketing suggests, but the items on it are genuinely valuable.

Technical audits at scale. Crawling thousands of pages, cataloging redirect chains, orphaned pages, duplicate titles, and broken canonicals, then drafting the fixes instead of just listing the problems. This is the strongest current use case because the work is verifiable. A redirect either resolves or it doesn't.

Internal-linking passes. Reading your corpus, identifying pages that should reference each other, and proposing the links with anchor text. Tedious for humans, mechanical for an agent, and easy to review in batch.

Metadata generation. Titles and descriptions for hundreds of pages against a template and character constraints. Review the batch, ship it.

Content refreshes against a brief. Give an agent a page, its current rankings, and a clear brief on what to update, and it can execute the refresh competently. The brief is doing the heavy lifting. The agent is doing the typing.

Programmatic page QA. Checking generated pages for thin content, missing schema, broken templates, and inconsistent formatting before they go live.

Monitoring-and-alerting loops. Watching rankings, crawl errors, and index coverage, then flagging anomalies with a diagnosis attached instead of a raw alert.

Notice the pattern. Every item is high-volume, mechanical, and checkable. The agent does work a diligent junior would do, faster, and a human confirms it before anything ships.

Where Do SEO AI Agents Fail?

They fail at judgment, and they fail quietly.

An agent cannot decide your positioning. It cannot tell you which keywords map to buyers versus tire-kickers, whether a comparison page should concede a competitor's strength, or what your product is genuinely best at. Those calls require knowing what's true about your business, and no crawl surfaces that.

Left unsupervised, agents optimize what they can measure and drift toward generic output. Word counts go up. Headers get keyword-stuffed into blandness. Every page starts sounding like the median page in your category, because the median is what the model learned. You end up with a site that scores well on an audit and says nothing.

The sharper risk is fabrication. An agent refreshing a product page will happily invent a statistic, a customer count, or a capability claim, and unlike a chatbot draft you'd read before sending, agent output is designed to ship. A hallucinated claim on a production page is a legal exposure and a sales problem, not a typo. This is why the review gate is the architecture, not a setting.

Why Does Context Matter More Than the Agent?

I made this argument about ChatGPT for marketing and it applies with more force here: generic inputs cannot produce differentiated output. An agent without your positioning, ICP, proof points, and approved claims doesn't produce bad work. It produces fast generic work, which is worse, because volume makes it look like progress.

Prompting at least has a human in the loop each time, filling in missing context from memory. An agent runs the same thin context through fifty pages before you look up. Whatever the agent doesn't know about your business gets averaged out of every page it touches.

So the prerequisite for agent-driven SEO isn't a better agent. It's encoded context: who you sell to, what you claim, what you can prove, how you talk, and what you never say. That's the same foundation that makes AI agents for marketing work beyond SEO — one governed context that every agent executes against, instead of each tool improvising its own version of your company.

Get the context layer right and a mediocre agent produces usable work. Skip it and the best agent on the market produces polished sameness.

When Should You Use an AI SEO Agent?

Use one when the work is high-volume and mechanical, the output is verifiable, and a human review gate sits between the agent and production. Technical cleanup on a large site. Internal linking across a few hundred posts. Metadata at programmatic scale. Refreshes where you've already decided what good looks like and written it into a brief.

Don't use one expecting it to build authority. Authority comes from original data, earned mentions, real expertise, and being cited by people who matter — none of which an agent can manufacture. And don't use one to decide strategy. An agent asked "what should our SEO strategy be" will return a competent summary of everyone else's.

A useful test before delegating any task: would you hand this to a smart junior contractor with zero institutional knowledge and a strict review process? If yes, an agent is probably cheaper and faster. If the task needs someone who knows the business, it needs you or someone who reports to you.

What Do AI Agents Change About SEO Itself?

The economics. When execution gets cheap, everyone's technical debt gets fixed, everyone's metadata gets optimized, and everyone's internal linking gets tightened. Execution quality stops being a differentiator because it stops being scarce.

What stays scarce: original data nobody else has, a point of view a model can't average into existence, and entity authority built over years of being the credible source on your subject. Those are the inputs agents can't generate, which makes them the inputs worth investing in.

Conveniently, they're also exactly what wins in AI search. The factors that get a brand recommended by ChatGPT — consistent entity signals, third-party corroboration, genuinely citable content — are differentiated-input problems, not execution problems. The agent handles the mechanical layer so you can fund the layer machines can't do.

Buy the agent for the drudgery. Budget the savings for the substance.

Frequently Asked Questions

What's the difference between an AI SEO tool and an AI SEO agent?

A tool assists with a single task you're driving: a crawl, a keyword report, a draft. An agent plans and executes a multi-step workflow toward a goal, with human checkpoints. Most products marketed as agents today are tools with a conversational interface. Ask the vendor to show you the workflow it completes end to end, and where the review gates sit.

Can an AI SEO agent replace an SEO team?

No. It replaces the mechanical portion of the work: audits, linking passes, metadata, refresh execution. Strategy, positioning, and judgment about what's true and provable about your business stay human. Teams that delegate judgment to agents ship confident, generic, occasionally fabricated pages faster than anyone catches them.

Are AI SEO agents safe to run on a production site?

Only with review gates. Agent output should land as drafts and pull requests a human approves, never direct-to-production changes. The failure mode isn't a broken site. It's a plausible-looking claim you never made, live on a page your buyers read.

Originally published at strategnik.com/thinking/ai-agents-for-seo.

AEO tracking tools measure how often AI assistants mention, cite, and describe your brand when buyers ask questions you care about. They work by running a fixed panel of prompts against ChatGPT, Perplexity, Gemini, and Google AI Overviews on a schedule, then comparing the answers over time.

That is the entire category. If you want a tool-by-tool review of the broader landscape, including schema, content optimization, and AI Overview rank tracking, read the full GEO and AEO tools roundup. This post covers the measurement layer specifically: what tracking captures, how it works mechanically, which metrics deserve a B2B marketing leader's attention, and when paying for monitoring is premature.

The short version of the point of view: tracking tools measure progress. They do not create it.

What Does AI Visibility Tracking Actually Measure?

Four things, and it helps to keep them separate because vendors bundle them differently.

Share of voice in AI answers. For a defined set of prompts, how often does your brand appear in the response? If a buyer asks an assistant for freight audit software and you show up in six of ten runs while a competitor shows up in nine, that gap is your share-of-voice problem stated plainly.

Citation tracking. Which URLs get cited when your brand appears, and whose pages are they? Sometimes the answer cites your own site. More often it cites third-party pages: review sites, industry publications, community threads. Knowing which sources power your mentions tells you which sources to protect and which to cultivate.

Sentiment and accuracy. How does the model describe you? An assistant can mention you and still misstate your category, your pricing model, or who you serve. A mention that positions you wrong can cost you the shortlist as surely as absence does.

Competitor benchmarking. The same prompt panel, run for your competitors. This turns raw mention counts into relative position, which is the only frame that matters in a competitive evaluation.

How Do AEO Tracking Tools Work Under the Hood?

Every tool in this category runs the same basic loop. Define a prompt set, query the answer engines on a schedule, parse the responses for brand mentions and citations, and diff the results against previous runs.

Understanding that mechanism changes how you should read the output. AI answers are generated fresh each time, so the same prompt can produce different responses hour to hour. What a tracking dashboard shows you is a probabilistic sample, not a ranking.

There is no "position 3" in ChatGPT. There is only a mention rate across repeated runs. Expect variance between readings, and treat any single data point with suspicion. The signal lives in trends: mention rate climbing over six weeks, a competitor's citation source dropping out, your description shifting after a site update.

Teams that miss this end up chasing noise. One bad weekly reading triggers a fire drill, and one good one gets screenshotted for the board. Both are the wrong response to a sampled measurement.

The other implication: the prompt panel is the product. A tool tracking fifty generic category prompts will produce clean charts about questions your buyers never ask. Before evaluating any vendor, write the fifteen prompts your actual evaluators would type into an assistant, because those are the only prompts whose answers you should pay to watch.

Which Metrics Matter for a B2B Team?

Most dashboards will show you more numbers than you need. Three matter.

Mention rate on buyer-intent prompts. Not "what is [category]" prompts. Prompts a real evaluator would ask: best tools for a use case, alternatives to an incumbent, comparisons between named vendors. Visibility on definitional prompts is nice. Visibility on shortlist prompts is pipeline.

Presence in shortlist answers. When an assistant produces a "top options" style response for your category, are you in it? These answers function like an analyst shortlist that regenerates on every query. Being consistently absent from them means the model does not consider you a default option, whatever your actual market position.

Cited-source overlap with competitors. Look at which sources power competitor mentions on prompts where you are absent. If three review sites and one industry publication keep feeding their visibility and you appear on none of them, that list of sources is your earn-list. This is the single most actionable output a tracking tool produces, because it converts a visibility gap into a concrete outreach and content plan.

Which Tools Do the Tracking?

Category-level guidance here, since the roundup linked above covers the landscape in more depth.

Profound sits at the enterprise tier: citation monitoring across generative answer surfaces, built for brands with the demand footprint to justify it.

Otterly.AI and Peec AI are mid-market AI visibility trackers, built around prompt-based monitoring of where your brand and competitors appear in AI answers.

Semrush has added AI visibility features to its broader suite, which is the shortest path for teams already living in its dashboards.

Pick based on where your reporting already happens and how much prompt-panel customization you need. The underlying tracking mechanics are broadly similar across the tier.

Should You Buy a Tracking Tool Before You Have Visibility?

Usually not, and this is where most teams get the sequence backwards.

Run a free audit first. You can do it manually in an afternoon: write ten buyer-intent prompts, run each several times across two or three assistants, and record mentions, citations, and descriptions. The full method is in how to check what ChatGPT says about your brand. Or use Strategnik's free AI Visibility Grader, which runs the structured version for you.

The audit tells you which situation you are in. If your mention rate is meaningfully above zero, monitoring makes sense. You have a baseline worth defending and trends worth watching. If your mention rate is zero, a dashboard will faithfully report zero every week while charging you for the privilege.

Zero-visibility companies do not have a measurement problem. They have an answer engine optimization problem: thin entity signals, no credible third-party citations, content that never answers the questions buyers actually ask assistants. That work has to happen first. Then tracking earns its keep by telling you whether the work is landing.

Sequence it this way: audit free, fix the visibility gaps, buy monitoring when there is movement worth measuring.

Frequently Asked Questions

How do I track AI search visibility for free? Build a fixed prompt panel of ten to fifteen buyer-intent questions, run each prompt several times across ChatGPT, Perplexity, and Gemini, and log mentions, cited URLs, and how you are described. Repeat monthly with the same panel so results stay comparable. The consistency of the panel matters more than its size.

What is a good AI mention rate? There is no universal benchmark, and anyone quoting one is guessing. The useful comparisons are relative: your rate versus named competitors on the same prompts, and your rate this month versus last. Direction and gaps beat absolute numbers.

Why do my AI visibility numbers change between runs? Because answers are generated fresh each time rather than retrieved from a fixed index. Tracking tools sample a moving target. Variance between individual runs is normal, and only sustained movement across multiple readings should change your decisions.

Originally published at strategnik.com/thinking/aeo-tracking-tools.

ChatGPT is genuinely useful for marketing. It turns a messy brief into a first draft, pressure-tests a landing page, summarizes a customer call, and lets a small team ship more variants than it could by hand.

And almost every team is using it in the way that produces the least value.

The difference is context. A model that doesn't know your positioning, ICP, proof points, claims, competitors, and voice is a fast intern who doesn't know your business. It produces plausible marketing. It sounds clean. It also sounds like everyone else running the same model on the same thin prompt.

What ChatGPT Is Actually Good At in Marketing

ChatGPT is good at production work — once the judgment has already happened.

First drafts and variants. Give it a real brief and one idea becomes ten headlines, three email angles, five ad versions, or a landing page draft worth editing.

Summarizing research and calls. Feed it customer interviews, sales notes, win-loss transcripts, or support themes and it pulls patterns faster than a person scanning the same pile by hand.

Repurposing across formats. A webinar becomes a recap, an enablement note, a social thread, a nurture email, and a product brief, when the source material is strong.

Critique. Ask it to red-team a page against a specific ICP, flag vague claims, or compare a draft against a positioning doc. That's usually worth more than asking it to write from scratch.

Structured brainstorming. Not "give me campaign ideas." That produces mush. But "given this audience, offer, constraint, and these claims, generate ten angles and score them on sales usefulness" gets you somewhere.

The pattern: ChatGPT works when it operates on something real. It fails when you ask it to invent the substance.

Where It Fails (And Why Your Output Sounds Like Everyone Else's)

The generic-output problem isn't mysterious. Most prompts are generic inputs asking for differentiated output. That math doesn't work.

If the model doesn't know your customers, it writes for a composite buyer. If it doesn't know your claims, it defaults to broad category language. If it doesn't know your competitive reality, it describes the market exactly the way your competitors do. If it doesn't know what you're allowed to say, it either overclaims or sands the message down until nothing sharp survives.

Your output sounds like everyone else's because your input looks like everyone else's.

The second failure is organizational. One person prompts ChatGPT with one version of the company. Another prompts it with a different one. Demand gen uses one ICP, content uses another, sales enablement has a third. Nobody's being careless. They're each filling in the missing context from memory.

That's how team-of-one prompting becomes five versions of the company.

Then the session ends. The next prompt starts from zero. The next teammate starts from zero. The next tool starts from zero. ChatGPT remembers within a conversation, and some products retain preferences, but that's not a governed operating context every teammate and agent shares.

The model isn't the problem. The blank slate is.

ChatGPT Prompts for Marketing That Actually Work

The best marketing prompts aren't magic words. They front-load the business context before asking for output.

"Here's our ICP, positioning doc, and the claim we're allowed to make: [paste]. Draft X."

Gives the model a boundary before it writes a word.

"Here's a customer call transcript and our ICP definition: [paste]. Extract the pains, objections, and exact language worth reusing."

Turns ChatGPT into a summarizer of evidence, not an inventor of insight.

"Here's our landing page and competitive framing: [paste]. Red-team it for vague claims and missing proof."

Uses the model for critique, where it exposes weak spots fast.

"Here's the source asset, our voice rules, and the target channel: [paste]. Repurpose it without adding new claims."

Keeps format conversion from becoming message drift.

"Here's our campaign brief and scoring criteria: [paste]. Generate ten angles, then rank them on fit, specificity, and sales usefulness."

Makes brainstorming structured enough to be useful.

Every one of these does the same thing: it hands the model context instead of a blank page.

From Prompting to a System

Prompting is per-person, per-session. That's the ceiling.

A good marketer gets better output than a bad one because they know what context to include. But that knowledge stays trapped in the person. The next teammate doesn't inherit it. The next session doesn't inherit it. The next agent doesn't inherit it.

A system encodes the context once.

That's the role of the Context Layer: your go-to-market context (persona, competitor, vertical, buyer stage, topic, product, channel, goal, and creative) encoded in a form machines can actually use, instead of pasted in from memory one prompt at a time. It isn't a prompt library, and it's no longer a bespoke consulting deliverable: the engine that tags that context and resolves which piece applies is open source (ontogent-core).

Once that layer exists, ChatGPT for digital marketing stops being isolated prompts and becomes an operating model. Agents draft against the same voice. Campaigns generate from the same ICP hierarchy. Content gets checked against the same claims and proof. The system doesn't depend on each person remembering which paragraph to paste.

That's also where the work moves from chatbot to fleet. A chatbot waits for a person to ask. A governed agent executes against context, constraints, and tools. The next step isn't a better prompt library. It's AI agents for marketing operating from shared context, along the path from chatbot to agent fleet.

ChatGPT is a useful interface. It isn't the system.

Frequently Asked Questions

Can ChatGPT replace a marketing team?

No. It replaces production drudgery, not judgment. Teams that treat it as a teammate with amnesia get burned — the model produces work faster than the organization can decide whether the work is right.

How do I use ChatGPT for digital marketing?

Start with context. Define the ICP, positioning, claims, proof points, competitors, and voice for the task. Paste the relevant source material instead of asking the model to invent from a topic. Then ask for a specific output, and a critique against your own criteria, before anything ships.

Is ChatGPT content bad for SEO?

Not because ChatGPT wrote it. Content is bad when it's thin, generic, unsupported, or unhelpful. Google's position is about usefulness and quality, not authorship. AI-assisted content works when it's grounded in real expertise, clear claims, and a coherent content system. Generic AI content fails because it has none of those.

Before you scale AI output, audit the context it's running on.

Originally published at strategnik.com/thinking/chatgpt-for-marketing.

The best GEO and AEO tools in 2026 are not magic visibility machines. The market is young, noisy, and full of vendors renaming ordinary AI SEO tools as generative engine optimization tools because the category is hot.

Most "AI SEO tools" are still content generators with a new label. That does not make them useless. It just means they should not be confused with a system for earning citations in AI answers, building entity authority, or making your site easier for machines to understand.

The tools that matter do one of four jobs: monitor AI visibility, improve entity and schema infrastructure, optimize content structure, or extend rank tracking into AI Overviews. Everything else is either workflow convenience or packaging.

The Four Jobs a GEO Tool Can Actually Do

Useful GEO and answer engine optimization tools fit four categories: AI-visibility and citation monitoring, entity and schema infrastructure, content structure and optimization, and rank or SERP tracking extended to AI Overviews.

The distinction matters because buying the wrong category creates false confidence. Monitoring tells you where you appear. Schema helps machines parse who you are and what you offer. Content optimization improves page-level clarity. AI Overview tracking shows whether search results are changing around your brand or category.

AI-Visibility and Citation Monitoring

This is the category most people mean when they go looking for the best-rated answer engine optimization tools: platforms that watch how brands, competitors, and sources appear inside AI-generated answers.

Profound is an enterprise-tier AI citation monitoring platform for tracking how brands appear across generative answer surfaces.

Otterly.AI monitors AI search visibility so teams can see where their brand, competitors, and content are showing up.

Peec AI tracks AI visibility and helps teams understand how they are represented in generative search results.

Semrush has added AI-visibility features to its broader SEO suite for teams that want AI search tracking alongside existing search workflows.

These tools are useful when you already have enough market presence to measure. They are less useful if the real problem is that your site has thin content, weak entity signals, or no credible third-party citations for models to rely on. For what these platforms measure, how the sampling works, and when buying one is premature, see AEO tracking tools: how to measure your AI search visibility.

Entity and Schema Infrastructure

Entity and schema infrastructure is the least glamorous part of GEO, but it is often where the work should start. Models and search systems need clear signals about your organization, products, authors, services, locations, relationships, and proof points.

This category includes schema markup generators, structured data testing tools, knowledge graph cleanup, organization schema, product schema, FAQ schema where appropriate, and content architecture that makes entities explicit. Even the best AI SEO tools cannot compensate for a site that makes your business hard to parse.

If you are serious about answer engine optimization, schema is not the whole strategy. It is the wiring. It helps machines understand the page, but it does not make the page authoritative.

Content Structure and Optimization

Content structure tools help teams make pages clearer, more complete, and easier to compare against the pages already ranking or being cited. This is where traditional SEO tooling overlaps with LLM SEO tools.

Surfer is a content optimization platform that helps teams structure pages around topics, terms, and competitive search patterns.

Clearscope does related work with a focus on topical coverage and page clarity.

These tools can improve the raw material. They cannot decide whether the page deserves to exist, whether the site architecture supports the topic, or whether the brand has enough authority to be trusted in an answer.

Rank and SERP Tracking Extended to AI Overviews

Classic rank tracking still matters because AI answers do not replace the entire search results page. They sit on top of, next to, or inside the same discovery environment buyers already use.

Ahrefs Brand Radar tracks brand mentions and visibility signals in AI Overviews so teams can monitor where they appear in AI-shaped search results.

Semrush also appears in this category because its AI Overview tracking extends traditional search monitoring into AI-generated SERP features.

This category is useful for teams that already watch rankings and want to understand how AI Overviews change visibility. It is not the same thing as full AI citation monitoring across every answer engine, and it is not the same thing as building authority.

What No Tool Does

No GEO tool builds entity authority for you. No AEO tool restructures your content architecture by itself. No monitoring dashboard earns citations from credible third-party sources. Tools measure and assist. They do not create the underlying reasons a model should trust, mention, or cite you.

That is system work. It includes clarifying entities, building high-intent content clusters, earning external references, improving technical structure, and aligning sales, product, and marketing language so the market describes you consistently.

If you need that system built, a GEO agency should be judged by the architecture it creates, not the dashboards it screenshots. The same is true for LLM SEO: the durable work is not prompt tricks. It is making your business legible across the web.

The deeper issue is usually not tooling. It is that the company has not built enough digital context for AI systems to understand why it belongs in the answer. That is why many B2B teams do not show up in ChatGPT even when they publish regularly. The problem is explained more directly in why B2B content does not show up in ChatGPT.

How to Choose

If you are an enterprise brand with existing demand, start with AI-visibility and citation monitoring. You need to know where you appear, which competitors are being cited, and which sources influence the answer.

If your site is structurally messy, start with entity and schema infrastructure. Better monitoring will only tell you that the foundation is weak.

If you already have a strong site but uneven content quality, use content structure and optimization tools to tighten pages, improve coverage, and remove ambiguity.

If your team already lives in SEO dashboards, add AI Overview tracking before buying a standalone platform. It is the shortest path from familiar search reporting to AI-shaped search visibility.

If you are choosing between categories, be honest about the job. Monitoring is not building. Optimization is not authority. Schema is not demand. The right tool depends on the bottleneck.

Frequently Asked Questions

What is the best AEO tool? There is no single best AEO tool because the category covers different jobs. Monitoring tools help you see whether you appear in answers. Schema tools help machines understand your site. Content tools improve page structure. None of those automatically builds the authority needed to be cited.

Are GEO tools worth it? Yes, if you know what you are buying. GEO tools are worth it when they answer a specific operational question: where do we appear, what are models citing, which pages are unclear, or how are AI Overviews changing the SERP?

Can I do GEO without tools? Yes. Start with clear content architecture, strong entity pages, schema, internal linking, and credible external references. Tools help measure progress, but the first layer of GEO is structure.

If you want to know whether your site gives AI systems enough context to trust you, start with a Digital Context Audit.

Originally published at strategnik.com/thinking/best-geo-aeo-tools.

For twenty years, SaaS go-to-market teams had a structural advantage: the product did the heavy lifting. Deep codebases, long roadmaps, compounding technical complexity, and expensive implementations meant that a strong feature set could create durable differentiation. Sales could lean on demos. Marketing could build campaigns around launches. Customer success could count on switching costs to protect retention.

That advantage is eroding. AI has compressed product development timelines to the point where a well-funded team can rebuild the functional equivalent of most SaaS products in months, not years. In some categories, it is faster than that. The feature you launched last quarter gets matched this quarter.

At Strategnik, we describe this using a physics framework. Features used to be a primary source of mass: the accumulated weight that creates gravitational pull toward your product and makes displacement expensive. When products were hard to replicate, that mass was intrinsic. When features become easy to replicate, each launch adds less mass. The gravitational pull weakens. Buyers drift more easily. Evaluations happen more frequently. Displacement gets cheaper.

But the same AI that erodes your product moat can build a different kind of moat faster than was possible two years ago. The teams that win the next five years are not the ones defending feature advantages. They are the ones using AI to build friction, brand gravity, and operational mass across the full go-to-market system.

And the infrastructure matters as much as the intent. Giving every team an enterprise license to Claude or ChatGPT does not get you there. A model without context is a fast intern who does not know your business. The system between the foundation model and the operator is what determines whether AI becomes a compounding advantage or an expensive autocomplete.

Two companies show how this plays out from opposite ends of the spectrum: Gong and Airtable. Gong built conversation intelligence through a top-down, sales-led motion. Airtable built a flexible database-spreadsheet hybrid that spread through organizations bottom-up, one team at a time. Both face the same structural threat from completely different positions.

Gong's core call recording and transcription is a commodity. Airtable's structured database with a clean interface is what every AI-native productivity tool is converging on. So if the features are replicable, where does the mass live? For Gong, it is the proprietary data: the organization's conversation history, deal patterns, coaching models, and forecast signals. For Airtable, it is the operational surface area: hundreds of bases across teams, each connected to automations, integrations, and workflows nobody fully maps and nobody wants to untangle.

The Sales Playbook

The old motion is breaking. A prospect can watch your demo on Tuesday, see an AI-native competitor ship something comparable on Wednesday, and ask on Thursday why they are paying a premium for a codebase that took a decade to build when the alternative took a quarter.

When your product carried intrinsic mass, discovery was about mapping the prospect's pain to your feature set. The implicit message was: "we built something you can't get elsewhere." That message worked because it was true. It is becoming less true by the month.

The shift is from selling capabilities to creating friction. The conversation has to move from what happens before implementation to what happens after. How deeply does this integrate into your workflows? How much proprietary data accumulates inside it? How painful would it be to unwind?

A Gong AE who leads with "we record calls, transcribe them, and surface deal insights" is competing on a feature checklist that AI-native tools are replicating monthly. Call recording is free. Transcription is a commodity API. Deal scoring is a feature every CRM is bolting on.

The winning pitch sounds completely different: "Your team has eighteen months of conversation data inside Gong. Every discovery call, every negotiation, every objection pattern. Your coaching models are trained on what winning looks like at your company, not in general. Replacing Gong does not mean finding a better call recorder. It means starting your conversation intelligence from zero."

That pitch is built on data mass, not features. In a data-driven sales org, losing eighteen months of accumulated pattern recognition is a cost that no feature comparison can offset.

Now layer AI onto that motion. A Gong AE using AI well does not walk into discovery with generic questions about sales process pain points. They walk in with a pre-built analysis of the prospect's revenue motion, pulled from job postings that reveal team structure and hiring velocity, earnings calls that surface sales productivity metrics, G2 reviews of the tools they currently use, and LinkedIn data that maps reporting relationships across the revenue org.

Airtable's sales motion is different because the product enters through PLG. By the time sales engages, the product is already embedded. The challenge is not getting the prospect to try it. It is converting grassroots adoption into an enterprise commitment before a competitor swoops in with a cheaper or shinier alternative.

The friction story lives in the operational graph already built: "Your marketing team runs their content calendar on Airtable. Your ops team tracks vendor management. Your product team manages the launch roadmap. Each base is connected to Slack, integrated with your data warehouse, and wired into automations that trigger downstream workflows. Replacing Airtable does not mean finding a better database tool. It means auditing, rebuilding, and reconnecting every workflow built on top of it across every team that adopted it independently."

AI makes this pitch surgical. An AI-equipped Airtable AE can analyze the prospect's existing usage data and generate an operational dependency map that visualizes exactly how embedded the product already is. That map becomes the enterprise sales narrative: "Here's what you've already built. Here's what it would cost to rebuild it somewhere else. Here's what it looks like when you formalize and expand it."

The AI-powered sales playbook has four shifts.

Pre-call intelligence at depth, for every account. The preparation that used to be reserved for six-figure deals can now happen for every qualified opportunity. AI agents can pull data from public filings, job boards, review sites, and API documentation to build a connection map before the first meeting. When every AE walks in with solutions-engineer-level context, discovery calls become architecture conversations. Architecture conversations produce implementation-heavy deals. Implementation-heavy deals create friction.

Custom proposals in hours, not weeks. AI can generate tailored implementation plans, ROI models, and integration architecture documents that used to take a solutions team days to build. When you deliver a custom-scoped proposal in the same week as the discovery call, you compress the sales cycle and shrink the window where competitors can insert themselves.

Competitive displacement playbooks that update in real time. Instead of static battlecards that are stale the week they ship, AI can monitor competitor product launches, pricing changes, and customer reviews continuously and surface relevant counter-positioning to AEs in the context of specific deals.

Pricing that adapts to the deal. As seat-based pricing loses its structural logic, sales teams need the ability to model consumption-based, outcome-based, and hybrid pricing for individual accounts. AI can model scenarios in real time during negotiations, grounded in the prospect's own publicly available financial data.

None of this happens by handing an AE a ChatGPT login. The company needs a connected data layer the model can query, pre-built workflows instead of blank prompts, competitive intelligence that feeds the model continuously, and guardrails that prevent hallucination in high-stakes contexts. Without that infrastructure, AI knows everything about the world and nothing about your business.

The Marketing Playbook

Most SaaS marketing organizations are still structured around product marketing as the center of gravity. Launch the feature. Write the blog post. Update the battlecard. Brief the analysts. Run the webinar.

That cadence made sense when features were durable. When the replication timeline compresses to weeks, the half-life of every announcement shrinks with it. You are constantly launching, constantly chasing, and the pipeline impact of each launch keeps diminishing. That is velocity without mass, and velocity without mass does not compound.

The shift is toward the things that do not commoditize. Features commoditize. Brand gravity does not. Distribution advantages do not. Proprietary data does not. The density of your presence across the channels and conversations where buyers make decisions does not. Marketing needs to reorient around these sources of mass, and AI is the tool that makes it possible to build them at the pace the market now demands.

Gong's durable mass is not the call recorder or the deal board. It is the dataset: millions of sales conversations across thousands of companies, plus the individual customer's own conversation history that grows more valuable with every call logged.

If marketing keeps leading with "we launched a new AI insight" or "our transcription is more accurate," they are competing on features every competitor is shipping simultaneously. The mass Gong should be marketing is harder to replicate: the proprietary data advantage that only exists because thousands of sales teams have been feeding their conversations into the platform for years.

AI changes the scale at which Gong can make that data advantage visible and authoritative. The marketing team can use AI to produce research-grade content derived from aggregated conversation data, content no competitor can replicate because no competitor has the dataset. That is not a feature announcement. That is a data moat turned into a content moat.

AI can also power real-time competitive response at a speed that used to require a war room. When a competitor launches a new AI feature and the LinkedIn discourse cycle starts, Gong's marketing should not be scrambling to draft a take. An AI-powered workflow can monitor the launch, draft a point-of-view piece grounded in Gong's proprietary data, route it through a human editor, and publish within hours.

Airtable's brand presence lives in a place most marketers would envy: the product distributes itself. One person signs up, builds a base, shares it with their team, and the tool spreads through the organization without marketing lifting a finger. The challenge is converting that passive distribution into active surface area before a competitor captures the same users with a shinier free tier.

AI makes this possible at a scale that changes the economics. Airtable's team can use AI to produce deep, specific workflow content for every use case emerging organically in their user base. Not generic "how to use Airtable" guides, but function-specific playbooks: how to build a content operations pipeline in Airtable that connects to your CMS, triggers Slack notifications for review cycles, and syncs with your analytics dashboard.

AI also enables Airtable to build and ship free tools that extend the brand's surface area into adjacent workflows: a project staffing calculator, an operations maturity assessment, a workflow automation audit that benchmarks a team's current process against best practices. Each tool captures data, generates leads, and reinforces Airtable's position as operational infrastructure rather than a spreadsheet alternative.

The PLG motion gives Airtable something else AI can amplify: user-generated social proof at scale. Thousands of teams are building workflows on the platform daily. AI can identify the most compelling use cases from product telemetry, generate case-study frameworks from usage patterns, and arm community managers with data-backed stories that turn organic adoption into marketing fuel.

The AI-powered marketing playbook, the core of a modern B2B SaaS marketing strategy, has four shifts.

Content velocity as a mass strategy. AI does not just make content faster to produce; it makes comprehensive coverage possible. For Gong, that means producing data-derived research content at a cadence no competitor can match. For Airtable, it means producing workflow-specific content for every use case, persona, and integration combination.

Answer engine optimization at scale. AI search is reshaping how buyers discover and evaluate software. The vendors whose content gets cited in AI-generated answers are the ones with the densest, most authoritative content corpus. If you are not in the AI-generated answer, you are not in the consideration set.

Free tools that compound brand gravity. Engineering-as-marketing used to require engineering resources that marketing could not access. AI has removed that constraint. Marketing teams can now build and ship diagnostic tools, calculators, and interactive assessments on their own.

Competitive monitoring that never sleeps. AI agents can track competitor launches, pricing changes, messaging shifts, customer reviews, and hiring patterns continuously, then surface actionable intelligence to the marketing team in real time.

Marketing needs breadth across the entire market and the ability to maintain quality at volume. The company needs a brand knowledge base the model works from, a content pipeline with human editorial gates, data connections to product and customer data, a multi-channel distribution layer, and an AEO layer built for how buyers actually find software now. AI-generated content at volume without editorial oversight is not a content strategy. It is a spam cannon.

The Customer Success Playbook

The traditional CS operating model is built around a simple assumption: if the customer is happy, they will stay. Measure NPS. Run QBRs. Resolve tickets quickly. Celebrate renewals.

That assumption held when switching was expensive regardless of satisfaction. A customer could be frustrated with your product and still renew because the cost of moving exceeded the cost of staying. In a world where switching costs are falling, satisfaction is necessary but not sufficient. A happy customer with shallow integration depth is one compelling demo away from evaluating an alternative.

The CS function has to shift from managing satisfaction to building mass inside each account, making your product so operationally embedded that the energy required to remove it exceeds anything a competitor is willing to subsidize.

Gong's mass is temporal and proprietary. Every call recorded, every deal tracked, every coaching insight generated adds to a dataset that belongs to that specific customer and cannot be transferred to a competitor. The conversation history is the moat, but only if the customer knows it.

A customer using Gong as a call recorder is one free Zoom transcription feature away from questioning the ROI. A customer whose sales managers run weekly coaching sessions based on Gong data, whose forecast depends on Gong deal signals, and whose competitive strategy uses Gong conversation trends has accumulated enough data mass that leaving means losing the institutional memory of how their team sells.

AI changes the CSM's ability to deepen that data mass. An AI-powered health model can analyze every account's usage patterns and generate a specific activation roadmap: "This team records calls but hasn't activated the coaching scorecards. Teams that activate coaching see a 23% improvement in new-hire ramp time. Here's the playbook to get their frontline managers running weekly coaching reviews in 14 days."

AI can also power proactive engagement at scale. When usage patterns shift, the system can flag the risk, generate a re-engagement message specific to that account, and recommend an intervention before the customer starts evaluating alternatives. Reactive CS waits for the churn signal. AI-powered CS catches the pre-signal.

Airtable's mass is structural and distributed. It does not accumulate in a central dataset like Gong's; it accumulates in the operational graph: bases spread across teams, each with their own automations, integrations, views, and workflows. The mass is not in any single base. It is in the aggregate, the unmapped web of dependencies nobody fully understands.

AI can close that visibility gap. An AI-powered CS system can analyze the customer's entire Airtable footprint and generate a dependency map that visualizes the operational surface area: "Your organization has 47 active bases across six teams, connected to Salesforce, Slack, and your data warehouse through 23 automations. Here's the dependency graph. Here's what breaks if this goes away."

AI also changes Airtable's CS expansion motion. Instead of generic recommendations to "try the enterprise plan," the system can identify specific expansion opportunities based on usage patterns. Every expansion recommendation that lands adds surface area. More bases, more connections, more mass.

The AI-powered customer success playbook has four shifts.

Predictive health scoring based on mass, not sentiment. Stop measuring account health by survey responses and support ticket volume. Measure it by structural indicators: active integrations, workflow dependencies, data volume, and user penetration across departments.

Automated expansion recommendations, account by account. AI can analyze each account's configuration, compare it against your highest-retention customer profiles, and generate a specific list of features, integrations, and workflows that would increase that account's mass.

Onboarding that builds friction from day one. AI-powered onboarding can detect the customer's tech stack, generate a custom integration plan, and guide the customer through setup in the first week, not the first quarter.

Churn intervention before the signal is visible. Traditional churn models wait for lagging indicators. AI can detect leading indicators: a competitor's sales team engaging the customer's LinkedIn connections, a decrease in integration utilization, or a shift in login patterns that suggests evaluation behavior.

CS has the most data-intensive infrastructure requirement in the GTM stack. The AI needs to understand the state of every individual account in real time. The company needs a live account health model connected to product telemetry, integration and workflow mapping per account, automated playbook generation, early warning systems with intervention triggers, and an implementation acceleration engine.

The Moat Is the System

Every integration scoped during a sales cycle is a root that deepens after close. Every piece of content that teaches the market how to think about your category makes the next buyer easier to reach. Every workflow dependency built by CS makes the account harder to unwind.

That is the new moat. Not a feature. Not a launch calendar. Not a QBR deck. The moat is the organizational system connecting sales, marketing, and customer success: sales scopes the friction, marketing builds the gravity, and customer success turns adoption into operational mass.

The foundation model is a commodity. Every company can buy access to Claude or GPT. What is not a commodity is the proprietary data layer, the brand knowledge base, the account-level telemetry, the workflow engineering, and the feedback loops that turn a general-purpose model into a GTM engine that only works inside your company.

Features will not hold the system together anymore. The companies that build the AI infrastructure to sell friction, market brand gravity, and retain through operational mass will compound while competitors are still defending feature checklists.

Originally published at strategnik.com/thinking/saas-playbook-features-not-moat.

The most capable model is the wrong default, and almost everyone is using it anyway.

Here's the move I see in every team that's adopted AI coding agents: they pick the smartest, most expensive model up front "to be safe," and never touch the setting again. It's a rational choice. Choosing a model per task is friction, and nobody wants the tool to stall on the one hard problem that actually mattered. So they over-provision. Now they're paying frontier prices to rename a variable, reformat a table, and fix a typo.

That's the whole bug, and it isn't really about code. It's a procurement habit: when triage feels like overhead, you match the most capable option to every task. It's flying every employee business class for the commute because sorting out who actually needs it is annoying. The cost doesn't track the value of the work. It tracks your unwillingness to decide.

The model should follow the work

The fix is obvious once you name it: spend should track the difficulty of the problem. Cheap, fast model for the mechanical 80%. Expensive, frontier model for the hard tail where reasoning actually changes the outcome. A capable mid-tier model for everything in between — and for deciding which bucket a given task falls into.

The catch, at least in Claude Code, is that a session pins one model start to finish. It can't downgrade itself once it's reading your prompt. So the only real mechanism for "use the right model for the job" is a router: a working model that reads each task, judges it, and hands the work off to the right specialist.

I built that router and open-sourced it. It's called overkill.

How it works

A capable mid-tier model runs by default. It has to read your prompt anyway — so judging difficulty is one beat of that same read, not a separate classifier call with its own latency and cost. As it reads, it routes:

• Mechanical, self-verifying work — renames, copy edits, lookups, boilerplate — goes to the cheapest fast model, at volume.

• Hard reasoning — architecture, optimization, the subtle bug that's eaten your afternoon — gets escalated to the frontier model.

• Everything else stays with the mid-tier model that's already doing the reading.

The triage is effectively free, and your bill starts to look like the actual shape of your work instead of a flat top-tier rate. Set it once; every project inherits it; any repo can override it.

That part is a cost story, and it's real. But it's not the part I'd defend hardest.

Newer is not better for the job

The reflex when a new frontier model ships is to make it the default immediately. It's the best one now, so it should run everything. That reflex is the same mistake in a new outfit — letting the most powerful, most expensive option become your floor because deciding is work.

overkill pins specific, tested model versions rather than floating to "whatever's newest." A release never silently takes over a tier. Every upgrade is a deliberate decision, and there's a ritual for it:

Read what the model is actually good at — the announcement, the model card, the stated strengths and the real price and latency, not the vibes. Map those strengths to the jobs you actually have: stronger reasoning is a candidate for the hard-tail tier; cheaper-but-still-capable might shift the default; a better cheap model upgrades the mechanical tier. Then decide the shape, not just the version number — sometimes a model is good enough to collapse two tiers into one, and sometimes the honest answer is to hold because the gain doesn't clear the cost of switching and re-testing. Run a few real tasks at each tier before you commit. Then log the decision and why.

The most uncomfortable line in that ritual: a better default can make a tier — or the whole router — unnecessary. If a stronger everyday model means you stop reaching for the frontier tier, that's not a failure of the system. That's the system telling you the truth. The scaffolding is a liability the moment the model outgrows it, and you want to be the kind of operator who notices.

The actual takeaway

Your AI spend is not a single switch. It's a portfolio of model-to-task decisions, and the default-to-the-best-one posture quietly makes every one of those decisions for you, badly. The win here isn't a cheaper model. It's the discipline of matching capability to task — and re-earning that match every time the lineup changes, instead of inheriting whatever's newest and most expensive as your standing default.

The tool is a hundred lines of config. The posture is the point.

overkill is MIT-licensed and on GitHub. Clone it, run the installer, start a session. Or just steal the idea: stop paying your smartest, most expensive resource to do your dumbest work.

Originally published at strategnik.com/thinking/stop-overpaying-for-ai-models.

If your marketing team still has a dedicated SEO person, a product marketing manager, and a production designer — you are paying three people to do work that AI does in minutes. Not poorly. Not "good enough." Well. The question isn't whether those roles will be absorbed. It's whether you'll restructure proactively or bleed budget until you're forced to.

Most CEOs I talk to know their marketing org feels wrong. They can't always articulate why. They just know they're spending $1.5M+ on a team of 8-12 people and the output doesn't match the investment. Pipeline is inconsistent. Content feels generic. Every initiative takes too long. The CMO asks for more headcount and the CEO can't figure out what the last three hires actually produce.

Here's what's happening: you're running a 2019 org chart in a 2026 capability environment. That chart was organized by channel — Paid Media, SEO, Content, CRM, Sales, Brand — with one specialist per lane. It was always a little broken. Each lane operated independently, measured different things, and created natural friction at every handoff. But it was the best structure available when each channel genuinely required deep, specialized knowledge to execute.

That's no longer true. And the gap between the old structure and what's now possible is where your money is going.

The Old Model Was Always Kind of Broken

Let me be honest about this, because the AI conversation tends to frame the old model as something that worked well and is now being disrupted. It didn't work that well.

Channel-based org charts created silos by design. Your SEO person optimized for rankings. Your content person optimized for volume. Your demand gen person optimized for MQLs. Your PMM optimized for sales enablement. Each of them reported their own metrics, hit their own targets, and declared victory — while pipeline stayed flat and CAC kept climbing.

The CEO saw the symptom: "marketing isn't working." The CMO's answer was always structural: we need another person. A lifecycle marketer. A growth person. A marketing ops hire to wire everything together. The team grew from 6 to 8 to 12, and the fundamental problem — that nobody owned outcomes, only channels — never got addressed.

This is the model you're defending when you resist restructuring. Not a well-oiled machine that AI is threatening. A Rube Goldberg apparatus that happened to be the only option when every channel required a human specialist to operate.

What AI Actually Changed

AI didn't just make marketing faster. It collapsed the skill premium on three entire specialties.

Product marketing used to be the translation layer between product and market. Positioning documents, competitive battlecards, sales enablement decks, launch plans — all of it required a person who could hold the product's technical reality in one hand and the buyer's decision framework in the other. That translation work was genuinely hard and genuinely scarce.

It's not scarce anymore. An AI agent with access to your product docs, win/loss data, competitive intelligence, and CRM produces a positioning draft, a battlecard, or a launch brief in minutes — work that used to take a PMM a week. Not a perfect draft. But a draft that's 80% there, which means the remaining 20% is judgment, not production. You don't need a $160K PMM to apply judgment 20% of the time. You need a strategist who evaluates and sharpens output as part of a broader role.

Design followed the same arc. The bottleneck in marketing design was never taste — it was production. Resizing assets, building landing page variations, creating social graphics, adapting templates for different channels. AI handles all of that now. The design skill that still matters is brand architecture: defining the system that governs how everything looks and feels. That's a creative director function, not a production designer function. One person plus AI replaces a team of three or four — and the output is more consistent because a system is governing it, not a queue of tickets.

SEO is the most dramatic collapse. Traditional SEO — keyword research, on-page optimization, technical audits, link building strategies — was a specialty because the rules were complex and always changing. Google's algorithm was a black box that rewarded people who studied it obsessively. Now AI tools run technical audits in seconds, generate schema markup, optimize content structure, and monitor rankings automatically. The strategic question — what should we be known for and how do we earn authority — was always a brand and content strategy question, not an SEO question. The tactical complexity just obscured that, and you paid $140K a year for the obscurity.

As a CEO, you should be asking: how many people on my marketing team are doing work that AI now does as well or better? Not "could theoretically do" — actually does, today, with commercially available tools. The answer is probably higher than your CMO has told you.

The New Structure: Organized by Outcome

The modern marketing org isn't organized by channel. It's organized by outcome — with architect layers underneath that make the whole system work.

Three outcome pods:

Acquisition owns the top of funnel. Paid, GEO/AEO, landing pages, conversion. This person thinks in pipeline creation and cost of acquisition, not channel-specific metrics. They deploy AI agents to manage the channels and focus their own time on strategy, creative differentiation, and conversion architecture. One person doing the work that used to require a demand gen manager, an SEO specialist, and a paid media buyer — because the execution complexity that justified three roles has been absorbed by tools.

Engagement & Monetization owns the existing customer. Retention, activation, expansion, lifecycle. This role used to be split across CS, lifecycle marketing, and product marketing — all measuring different slices of the same thing: how much value does this customer extract and how do we compound it? Combining them into one owner with AI-powered automation eliminates the handoff gaps where customers quietly churn.

Brand owns brand in market. Creative direction, production, voice. Not brand in the old "awareness campaign" sense — brand as the system that makes everything the company produces feel like it comes from the same mind. This is the role that becomes more important as AI generates more first drafts, because without a governing system, you publish 10x the content at 0.1x the coherence.

Three architect layers that sit across the entire org:

Orchestration — the person or small team that owns workflows, agents, and automation. This function didn't exist in 2019 because there was nothing to orchestrate. Each channel ran its own playbook. Now the playbooks are interconnected, the agents are cross-functional, and someone needs to design and maintain the system that connects acquisition signals to engagement actions to brand consistency. I've written about this as the Chief Marketing Orchestrator role — it's the single most important new function in marketing.

Creative and Voice System — the architect layer that owns brand guidelines, messaging framework, voice and tone, and ICP definition. This isn't a person who makes ads. This is the person who builds the system that ensures every piece of output — whether generated by a human or an AI agent — sounds like the same company. Think of it as the constitution that every agent and every team member operates under.

Data — owns infrastructure, decisioning, and signal. Not analytics in the old "here's a dashboard" sense. Data as the nervous system that tells the pods what's working, what's changing, and what to do next. This person builds the feedback loops that make the outcome pods self-correcting rather than reactive. Without this layer, you're flying the new structure with the old instruments.

The Math CEOs Should Run

A 2019 marketing team at a Series B company typically ran 8-12 people. PMM, SEO specialist, content writer, designer, demand gen manager, lifecycle/CRM person, brand marketer, marketing ops, a couple of junior hires filling gaps. All-in cost: $1.2M-$1.8M in salary alone, plus tools, plus management overhead.

The 2026 version of that team — producing more output, with better targeting, across more channels — is 4-6 people plus an agent fleet. Three pod leads, an orchestration architect, and one or two specialists in the architect layers depending on where the company's complexity lives. All-in cost: $800K-$1.1M in salary, plus $50-100K in AI tooling.

That's not a cost-cutting story. It's a leverage story. You're not spending less. You're getting dramatically more per dollar — and the "more" shows up where it matters: in pipeline, in conversion, in customer expansion. Not in activity metrics that make the team feel busy.

But here's the part most CEOs miss: the savings aren't the point. The point is that the new structure is simply better at producing outcomes. Even if it cost the same, you'd want it. The channel model creates handoff friction, metric fragmentation, and strategic incoherence by design. The outcome model eliminates all three. The fact that it also costs less is a bonus, not the thesis.

Why Your CMO Might Resist This

I want to be direct about this because it's the conversation most CEOs need to have and most consultants won't touch.

Your CMO may resist restructuring for legitimate reasons or illegitimate ones. The legitimate reason is that the transition is genuinely hard — you're asking people to change roles, absorb new tools, and operate at a higher altitude, all at the same time. That's real change management and it deserves real support.

The illegitimate reason is that the old structure is easier to manage. Six direct reports, each running their own channel, each reporting their own metrics — that's a coordination job, not a strategy job. Some CMOs built their careers on coordination. The new structure requires a CMO who thinks in systems, designs feedback loops, evaluates AI output against market reality, and treats marketing as an engine, not a department. That's a harder job. Not every CMO wants it.

You need to know which conversation you're having. If your CMO says "we're not ready for this" and follows it with a transition plan, you have a leader who gets it. If your CMO says "AI can't replace my team" and follows it with a defense of the status quo, you have a leader who's protecting a structure, not building one.

What This Actually Looks Like in Practice

The team that replaces your 12-person department isn't a skeleton crew. It's an architect and three agentic operators who produce at a level your current team can't touch — not because your people aren't talented, but because the structure they're operating in won't let them.

The architect is the person who redesigns the entire marketing engine. They define the outcome pods, build the orchestration layer, design the agent workflows, and establish the creative and voice system that governs everything. This isn't a consultant who hands you a deck and leaves. This is someone who rewires how your marketing function operates — the data flows, the feedback loops, the decision frameworks — and then runs it.

The three operators each own an outcome domain. The acquisition operator runs paid, GEO/AEO, landing pages, and conversion — not by doing the manual work, but by directing AI agents that handle execution at scale while they focus on strategy, creative angles, and conversion architecture. The engagement operator does the same for retention, activation, and expansion. The brand operator governs creative direction and the voice system.

Four people. Agent-augmented. Producing more coherent, higher-quality output across more channels than your team of 12 — because every hour of human effort is spent on judgment and strategy, not production.

This isn't theoretical. This is how the best marketing teams are operating right now, and the gap between them and the channel-based model is widening every month.

So What

If you're a CEO reading this, here's the test. Pull up your marketing org chart. Count the number of people whose primary value is channel execution — the SEO specialist, the production designer, the PMM who spends most of their time on decks and battlecards. Now ask: what would these people do if AI handled 80% of their production work?

If the answer is silence — if you can't articulate what strategic contribution they'd make without the execution work — you don't need more people. You need an architect who can rebuild the machine.

The model was always kind of broken. AI just made it indefensible. And the companies that figure this out first aren't going to win by a little — they're going to operate at a level that makes the old model look like it belongs in a museum.

Strategnik builds this. Let's talk about what your team should look like.

Nick Talbert is the founder of [Strategnik](https://strategnik.com), where he helps B2B companies design and execute go-to-market strategies built for how markets actually work. Previously, he held senior marketing leadership roles at numerous B2B technology-forward companies. [See full experience](https://strategnik.com/experience).Strategnik, where he helps B2B companies design and execute go-to-market strategies built for how markets actually work. Previously, he held senior marketing leadership roles at numerous B2B technology-forward companies. See full experience.*

Originally published at strategnik.com/thinking/marketing-org-chart-is-being-rewritten.

There's a post circulating from an outbound team that just rebuilt their entire context pipeline. Not because it was broken — because it worked too well.

They started with ten context files per client. Generic ICP, catch-all personas, high-level positioning. The simplest approach worked: dump everything into the prompt, cache the static parts, batch the calls. Cost low, latency irrelevant.

Then they got serious. They processed a hundred sales call transcripts through Claude, semi-automatically. Each call produced structured fragments — ICP refinements, persona nuances, specific objections they'd actually heard, proof points grounded in real conversations. They went from a generic "Head of Product" persona to specific buyer segments like "Head of Product at Early Stage UGC Platform."

The fidelity jumped. And the dump-everything pattern broke.

A hundred context fragments don't fit in a useful prompt. Even with aggressive caching, signal gets diluted and costs spike. So they built two new layers: a router that selects which subset of context is relevant for a given lead, and a composer that traverses the reference graph between notes — objection leads to proof, JTBD leads to persona, persona leads to insight — and bundles only what matters.

It's a clean solution. They're rightfully proud of it. And the pattern they discovered is the most important infrastructure problem in AI-powered GTM right now.

Every team building with AI agents will hit this wall. Most will hit it one client at a time and build one-off solutions. That's expensive.

Three stages every team goes through

The progression is predictable enough to name the stages.

Stage 1: Sparse context. You have a handful of documents — positioning deck, ICP overview, maybe a brand guide PDF. The constraint is generating anything useful at all. The fix is straightforward: put everything the model needs into the prompt. Every agent gets the full context dump. It works because there's not much to dump.

This is where most companies are today. They gave their team AI access and called it a strategy. The agents produce generic output because the context is generic. But at least it fits.

Stage 2: Dense context. Something changes. Maybe you process call transcripts. Maybe you accumulate six months of competitive intel. Maybe a thorough onboarding extracted twenty personas, forty objections, eight competitor displacement narratives, and a hundred proof points. The context is rich. Specific. Grounded. And it doesn't fit in a prompt anymore.

The team that wrote the original post got here through manual work — a human reviewing each transcript, refining each fragment, promoting the good ones. That's legitimate. It's also the path that doesn't scale to client two.

Stage 3: Routing becomes the bottleneck. You stop tuning the prompt and start architecting retrieval. The question shifts from "how do I give the model enough context" to "how do I give the model the right context." This is a fundamentally different engineering problem. It requires structure — typed relationships between context fragments, authority levels that distinguish confirmed knowledge from hypotheses, and a routing layer that can explain why it selected what it selected.

The original post nails this: "When context is sparse, the bottleneck is generating good content. When context is dense, the constraint is routing and distilling to the right content."

Correct. And predictable. Which means you can architect for it from the start.

Why bespoke solutions cost you twice

Here's where the pattern gets expensive. Every team that hits Stage 3 builds their own version of the same two components: a router and a composer. They build it for their specific use case (outbound sequences, in the original post's case). It works. Then they add a second use case — maybe content production, maybe ABM personalization — and discover that the routing logic is different for each one.

The outbound router selects context by buyer segment (persona × ICP intersection). The content router selects by topic pillar and buying stage. The ABM router selects by account-level signals and competitive displacement angle. Three use cases, three routers, three maintenance surfaces. The code that routes context for outbound doesn't know how to route context for a blog post. The composer that bundles objection-to-proof chains for a sales email doesn't know how to bundle competitive framing for a case study.

What started as a clean solution becomes a context routing sprawl. Each new use case adds a new routing layer. Each new client adds a new set of context fragments with different structures. The team that built this for one client now has to decide: rebuild it for client two, or abstract it into a platform?

Most choose neither. They copy-paste the first client's setup, modify it until it sort of works, and accumulate technical debt that shows up six months later as "why does our AI produce better content for Client A than Client B?"

The answer is always the same: the routing works for A because someone built it for A. It fails for B because nobody did.

What routing as infrastructure actually looks like

The Context Layer we've been building at Strategnik was designed for Stage 3 from the start — not because we're prescient, but because the Afiniti engagement taught us what happens when context gets dense fast.

Three things make the difference between a bespoke router and routing infrastructure.

Typed relationships, not flat documents. Every context fragment in the Context Layer has typed edges to other fragments. A forbidden term links to the brand voice spec it protects. An ICP links to the personas that serve it. A competitive disposition links to the proof points that support the displacement narrative. The router doesn't search through documents — it traverses a graph. That's the difference between handing someone a filing cabinet and handing them a map of how everything connects.

Authority hierarchy, not binary inclusion. Not all context is created equal. A positioning statement confirmed by the founder carries more weight than one inferred from website copy. The Context Layer encodes this as an authority hierarchy — codified, working, hypothesis, inferred — and the routing layer uses it to decide not just what to include, but how much to trust it. When the router selects a competitive framing node at hypothesis authority, it flags it: "this displacement narrative is inferred, not confirmed. Use with caution." The team that built the bespoke router treats all context fragments as equal. Ours doesn't.

Composability across use cases. The same graph serves outbound, content production, ABM personalization, and compliance validation — because the routing is parameterized by task type, not hard-coded per use case. An outbound composition request routes through persona × ICP intersection to select the buyer segment, then traverses objection → proof and pain → solution edges. A content composition request routes through topic pillar × buying stage to select the relevant cluster, then traverses entity → definition and claim → evidence edges. Different paths through the same graph. One infrastructure investment.

This is what the original post's author will build next — or should. The bespoke router they built for outbound is the right solution at the wrong layer. It belongs in the platform, not the pipeline.

The acquisition problem nobody talks about

The original post describes how they got to dense context: process a hundred call transcripts through Claude, semi-automatically, with a human reviewing and approving each one. That's the right methodology. It's also a hundred hours of operator time per client.

The context routing problem and the context acquisition problem are twins. Dense context creates the routing challenge. But acquiring dense context creates the scaling challenge. If every new client requires a human to review a hundred transcripts before the system produces sharp output, you don't have a platform — you have a consulting engagement with good tooling.

The answer is the same for both: encode the methodology into infrastructure. Automated extraction pipelines that pull structured context from a company's website, competitive landscape, customer data, and market signals. Semi-automated review flows that present hypotheses for human confirmation instead of requiring humans to generate from scratch. Completion checklists that verify the graph is dense enough to route well before declaring a tenant live.

The team that goes from "we manually built a hundred context fragments for one client" to "the system automatically discovers and structures context for any client" has something worth selling. The team that stays at "we do it manually and it's really good" has a services business.

Both are fine. But they're different businesses with different ceilings.

Where this is going

The original post ends with a prediction: "I think this will keep happening as agents accumulate institutional knowledge. The first thing you build is 'load the context.' Eventually you have to build 'find the right context.'"

I'd go further. The next bottleneck after routing is validation — how do you know the context you routed is still accurate? Competitive landscapes shift. ICPs evolve. Proof points expire. A routing layer built on stale context is worse than no routing at all, because it delivers wrong answers with high confidence.

After validation comes composition governance — not just "did we select the right context" but "did the output comply with the rules encoded in that context?" The forbidden term that should never appear in customer-facing content. The competitive framing that's approved for sales but restricted from public marketing. The proof point that's valid for one vertical but inapplicable to another.

Each of these — routing, validation, governance — is a component of what we call the Context Layer. Each one is a problem that every team hits in sequence as their context matures. And each one is dramatically cheaper to solve at the platform level than to rebuild per client.

The post that inspired this piece is a field report from Stage 3. It's accurate, well-observed, and correctly identifies the inflection point. The question for every GTM team reading it isn't "should we build a router?" — you'll have to eventually. The question is whether you build it as a one-off, or as the foundation for everything that comes after.

Originally published at strategnik.com/thinking/context-routing-bottleneck.

A company I advise was running paid campaigns on LinkedIn, Google, and Meta. They had a podcast, a blog, a monthly webinar series, and a Slack community. They sponsored two industry conferences per quarter. Their marketing team called it "full-funnel, multi-channel coverage."

Their pipeline was flat.

When we dug in, the problem was obvious. They were present in eight channels and memorable in zero. Every blog post restated the same positioning. The podcast was interview-format filler with no editorial point of view. The webinar replayed the demo. They had surface area in the most literal sense — they were spread across a lot of surface — but none of it carried weight. The market saw them everywhere and retained nothing.

That's not a surface area strategy. That's a budget allocation plan with extra steps.

What Surface Area Actually Means

In physics, surface area is how much of an object is exposed to its environment. More surface area means more interaction with surrounding forces — more heat exchange, more chemical reaction, more contact.

In B2B GTM, surface area is how much of the market you're in contact with. Not "how many channels are you on" but "where does your brand exist in the environments where your ICP works, learns, evaluates, and buys?" The distinction matters. Channels are plumbing. Surface area is presence at the points of transaction and decision.

The Coherence Model framework treats surface area as one of six forces governing growth. And like the physics concept, it doesn't operate in isolation. Surface area without mass is exposure without substance. Surface area with mass is how you build compounding authority across the entire buyer journey.

The Four Layers

Surface area isn't monolithic. It exists in four distinct layers, and most companies only think about one of them.

Content surface area is the most obvious: blog posts, guides, research, documentation, videos, podcasts. Each piece is a discoverable touchpoint. The test isn't how much content you produce. It's whether a buyer searching for any problem in your category — not just your brand name — finds you. Comprehensive content surface area means every conceivable search query in your space returns something from you. Thin coverage means you're invisible until someone already knows your name.

Distribution surface area extends your reach through structures you don't own. Integrations, technology partnerships, marketplace listings, referral networks, co-marketing relationships. A company listed in 1,600 integrations directories appears in buying conversations it never initiated. A company with three integrations appears only when it shows up under its own power. Distribution surface area is earned presence — the market surfaces you without you asking.

Community surface area is where your brand exists in conversations you don't control. Events, forums, Slack groups, LinkedIn comments, user communities. The strongest version of this is user-generated: customers recommending you in peer conversations, answering questions about your category with your name attached. You can't manufacture this, but you can create the conditions for it by building a product and brand worth talking about.

Product surface area applies to companies with usage-driven distribution. Every Calendly link in an email, every Loom video in a Slack channel, every Notion doc shared with a partner — these are brand impressions that compound with adoption. The product becomes its own growth channel. For PLG companies, product surface area is often the most powerful layer because it scales with the customer's own activity, not your marketing budget.

Surface Area and the Context Field

Here's where most people confuse two related concepts. Your context field determines how you show up at each touchpoint — your positioning, your authority, the weight your brand carries when someone encounters it. Surface area determines where you show up. How many touchpoints exist. How many contexts contain your brand.

You need both, and the sequence matters. If you expand surface area before you have a context field worth projecting, you're amplifying nothing. Every new channel broadcasts the same empty message to a wider audience. But if you build mass and a strong gravitational pull, then refuse to expand surface area, you're the best-kept secret in your market. Deeply credible to the twelve people who found you; invisible to the four thousand who should be buying from you.

The play is: build enough mass to be substantive at each touchpoint, then systematically expand surface area into every context where your ICP transacts, learns, and decides. Let the context field do the work at each point of contact.

The Trap: Surface Area Without Mass

This is the failure mode I see most often, especially in venture-backed companies with pressure to show "brand awareness" metrics.

The symptoms are easy to spot. Forty-seven blog posts that all say roughly the same thing in different words. Six social platforms with sporadic posting and no real engagement on any of them. Conference sponsorships where the booth staff can't articulate the differentiation. A podcast that's been running for eighteen months with flat listenership because every episode is a generic industry conversation that three hundred other podcasts are also having.

Wide reach. Zero weight. The market encounters you everywhere and forms no impression worth acting on.

This is worse than having narrow surface area, because it trains the market that you're generic. The buyer who sees you in ten places and learns nothing new in any of them has now categorized you as background noise. That impression is hard to reverse. You've spent money and time to become forgettable at scale.

How AI Changes the Surface Area Equation

Here's the part most B2B marketing teams haven't internalized yet. AI search — ChatGPT, Perplexity, Gemini, Claude, AI Overviews — doesn't work like traditional search. It doesn't return ten blue links ranked by page authority. It synthesizes answers from across the entire surface of the internet.

That means surface area is now a prerequisite for AI visibility. If your brand exists in only two or three contexts, the AI has limited signal to draw from. If your brand appears across content, documentation, community discussions, partner ecosystems, and earned media, the AI has a dense web of references to synthesize. More surface area, assuming it carries mass, means more raw material for AI systems to cite.

This collapses a wall that used to exist between channels. In the old model, your blog and your conference sponsorship and your integration marketplace listing operated semi-independently. In the AI search model, they're all inputs to the same synthesis engine. A prospect asking ChatGPT "what's the best platform for X?" gets an answer assembled from everywhere you've ever shown up. Or everywhere you haven't.

Surface area with substance is now the price of admission to AI-mediated discovery.

The Diagnostic

When I work with clients on surface area, we start with three questions adapted from a framework I've used with companies like Afiniti:

Where is the market transacting? Not where you wish they were buying, but where they actually evaluate, compare, and purchase solutions in your category. Marketplaces? Direct? Through partners? Through consultants? If you're not present where the transaction happens, the rest of your surface area is awareness without conversion.

Where is the market talking? Which communities, forums, conferences, and peer networks do your buyers participate in? Where do they ask for recommendations? Where do they share warnings? If you're absent from these conversations, your competitors are filling the gap with their narrative.

Where is the market deciding? What does the last mile of their evaluation process look like? Do they check G2 reviews? Ask their network on LinkedIn? Read analyst reports? Search for comparison content? The decision context is where surface area has the highest conversion leverage. Being present there with credible, substantive content is worth more than being present in fifty awareness channels.

Map all three. Then honestly assess: are you present at each of those points with substance, or are you missing from the places that matter most?

So What

If your pipeline growth has plateaued despite steady content production and marketing spend, the problem might not be your message or your budget. It might be that you're covering the wrong surface area — present in channels that don't map to where your ICP actually transacts, talks, and decides.

Run the diagnostic. Map where your market is actually active. Compare that map to where your brand shows up. The gaps are your growth opportunities, and the filled-in areas without mass behind them are your efficiency drains.

And if you're not thinking about AI search as part of this equation, you're building surface area for a discovery model that's already being displaced. The companies that show up in AI-mediated answers in 2027 will be the ones that built substantive, distributed surface area in 2026. The window to build that foundation is now.

Surface area is a growth strategy, not a channel plan. Treat it accordingly.

Nick Talbert transforms how B2B companies go to market. As founder of [Strategnik](https://strategnik.com), he helps founders and revenue leaders turn scattered tools and teams into pipeline, conversion, and growth that compounds.Strategnik, he helps founders and revenue leaders turn scattered tools and teams into pipeline, conversion, and growth that compounds.*

Originally published at strategnik.com/thinking/surface-area-for-growth.

You gave five teams access to AI. Each one produced content last week. Read all five pieces back to back and tell me if they sound like the same company.

I'll save you the trouble. They don't. One sounds like a blog post from 2019. Another reads like a press release written by committee. The third uses competitor language your positioning explicitly avoids. The fourth nails the tone but targets the wrong buyer. The fifth is genuinely good — and completely disconnected from the other four.

This is what scaling AI without shared context looks like. You didn't get more output. You got more fragmentation, delivered faster.

The Physics Metaphor (and Why It's Not a Metaphor)

In the Coherence Model framework, every company accumulates mass — brand credibility, market presence, accumulated authority. Mass is real. You can feel it in the difference between a cold outbound email from a company nobody's heard of and the same email from a company the prospect already respects.

In physics, mass generates gravity. Not as a choice. As a consequence. Mass bends the space around it, creating an invisible field that tells every object in the system how to move. Planets don't decide to orbit the sun. The gravitational field makes orbit the only coherent path.

Your company's equivalent is what I call the Context Field — the structured operating context that tells every tool, agent, team member, and channel how to behave. Not through policies people forget. Not through brand guidelines gathering dust on a shared drive. Through encoded context that is present every time a decision gets made, whether a human or a machine is making it.

Without a gravitational field, objects drift. They move in random directions, collide with each other, or fly off into space. The same thing happens to your GTM motion when five teams, eight agents, and twelve channels all operate without a shared foundation. Everyone is working. Nobody is coherent.

Why Most Companies Skip This

The gap is predictable. Most companies go from "we know who we are" straight to "let's be everywhere" — and they skip the step in between.

That step is encoding what you know into a system.

The founders know the voice. The head of marketing knows the ICP. The competitive intelligence lives in one person's head or a Slack channel nobody searches. The content strategy exists as institutional memory, not documented architecture. And the measurement framework is a spreadsheet that gets updated when someone remembers.

This works at small scale. The founding team carries the context in their heads, and proximity keeps everyone roughly aligned. Then you add AI tools. You add agencies. You add new hires who weren't in the room when the positioning was debated for three months. You expand to new channels. And suddenly the context that used to travel through hallway conversations doesn't travel at all.

The field was never built. It was implied. And implied fields don't scale.

What the Context Field Contains

The Context Field has six components, each encoded in machine-readable format — structured context that agents reference autonomously, not a PDF people forget to check.

Brand + Voice Spec. Tone, editorial standards, banned language, writing style rules. This is what makes five different agents produce output that sounds like one company.

ICP + Buyer Context. Ideal customer profiles, buying committee maps, pain hierarchy by persona, objection framing. This is what makes outreach targeted instead of generic.

Competitive Framing. Positioning against alternatives, displacement narratives, win/loss context, differentiated value claims. This is what makes content take a position instead of hedging.

Content Architecture. Topic cluster design, semantic entity relationships, atomization frameworks, internal linking logic. This is the structural blueprint that makes every piece of content reinforce every other piece.

Machine Readability + Distribution Schema. JSON-LD structured data, entity definitions, llms.txt, AI crawler accessibility, citation architecture. This is what makes your company legible to the machines your buyers consult before they ever contact sales.

Measurement Targets. Revenue influence metrics, conversion benchmarks, feedback loops. This is the steering mechanism that tells the system whether it's working and what to adjust.

I wrote a detailed breakdown of each component, what built looks like versus missing, and why you can't skip any of them: The Six Components of a Context Field.

How the Field Connects to the Rest of the System

The Context Field doesn't exist in isolation. It's the connective tissue of the entire Coherence Model framework.

Mass generates the field. Your accumulated brand credibility, market presence, and content authority create the raw material. The Context Field is what that mass actually does — it bends the behavior of everything around it toward coherence. A company with significant mass but no encoded field is a star that generates heat but no orbital system. Lots of energy, no structure.

The field extends across [Surface Area](https://strategnik.com/coherence-model/surface-area).Surface Area.** Every channel, every touchpoint, every agent-generated piece of content operates within the field — or it doesn't. Surface area without a context field is random drift at every touchpoint. You're present everywhere and coherent nowhere. The field is what makes wide reach compound instead of fragment.

A strong field reduces [Friction](https://strategnik.com/thinking/friction-problem).Friction.** When every team member and every agent works from the same encoded context, people stop wasting time re-deriving first principles. The SDR doesn't need to ask marketing how to position against Competitor X. The content agent doesn't need a briefing on buyer persona pain points. The agency doesn't need a three-week onboarding to understand the brand voice. The context is in the field. The friction disappears.

The field compounds [Momentum](https://strategnik.com/coherence-model/momentum).Momentum.** Each month the Context Field runs, it gets sharper. Measurement targets feed back into ICP context. Win/loss data refines competitive framing. Content performance sharpens the content architecture. The field learns. And because every agent and team member draws from it, the entire system gets better simultaneously — not one piece at a time.

Field Strength as a Diagnostic

You can diagnose a company's field strength in about thirty minutes. The symptoms are obvious once you know what to look for.

Weak field: Content sounds different across channels. New hires take months to produce on-brand work. Agencies need constant correction. AI tools produce output that requires full rewrites. Competitive positioning shifts depending on who's presenting. The sales deck says one thing, the website says another, and the SDR's email says a third.

Strong field: A new agent or team member produces coherent output on the first attempt because the context is encoded and accessible. Content across channels reinforces the same positioning without anyone coordinating it manually. Competitive framing is consistent whether it shows up in a blog post, a cold email, or a case study. The company sounds like one company, everywhere, every time.

The difference isn't talent. It isn't effort. It's infrastructure. A strong field makes average performers look coordinated. A weak field makes excellent performers look incoherent.

If you want to assess where your field stands, the Digital Context Audit is the diagnostic I run with clients to map the gaps.

The Compounding Advantage

Here's what most companies underestimate: the Context Field is a compounding asset. Every month it operates, it captures more signal, refines more context, and makes every downstream activity more efficient.

The company that builds this first in a competitive market gets an advantage that widens over time. While competitors are still briefing agencies, correcting AI output, and reconciling inconsistent messaging across channels, the company with a functioning Context Field is running a system where every new piece of content, every new agent, every new channel automatically inherits the full context of the organization.

This isn't a one-time project. It's infrastructure that appreciates. The gap between a company with a twelve-month-old Context Layer and one that just started isn't twelve months of content. It's twelve months of compounded learning, refined targeting, and accumulated coherence that no amount of sprint work can replicate.

The companies that figure this out in 2026 will look back on it the way we look back on companies that invested in SEO in 2010. Not because the tactic was novel, but because they built the infrastructure while everyone else was still debating whether it mattered.

So What

If you're adding AI tools, expanding channels, or scaling content production, the bottleneck isn't the tools. It's the operating context those tools work from.

Before you buy another platform, hire another agency, or deploy another agent, ask one question: does a shared, encoded, machine-readable operating context exist that tells every part of our system who we are, who we're talking to, why we win, and how to sound like us?

If the answer is no, you're scaling the chaos. More tools will produce more output. The output won't cohere. And you'll spend more time correcting the fragmentation than you saved by automating the production.

Build the field first. Everything else orbits it.

Nick Talbert transforms how B2B companies go to market. As founder of [Strategnik](https://strategnik.com), he helps founders and revenue leaders turn scattered tools and teams into pipeline, conversion, and growth that compounds.Strategnik, he helps founders and revenue leaders turn scattered tools and teams into pipeline, conversion, and growth that compounds.*

Originally published at strategnik.com/thinking/gravity-field-operating-context.