GEO became a checklist and missed the retrieval
Meta description: A practical AI visibility program joins technical SEO, original content, brand presence, competitive analysis and revenue measurement.
I have watched GEO arrive as a discipline in real time, and I have watched it collapse into a checklist in the same motion. The pattern in the companies I see: a team is handed the acronym, a vendor deck supplies the tactics, and within a quarter the company has FAQ schema on every page, a set of “definitive guides,” and a tool that scores its “AI readiness.” Meanwhile the engines cite a competitor’s narrow page that plainly answers a specific sub-question. The company is doing GEO. It is not being retrieved.
When a company asks me for an AI visibility strategy, I do not begin with a fashionable acronym. I begin with the business and the information already available. What is the site doing in conventional search? Where is its traffic coming from? Which pages bring inquiries? Which competitors occupy the strongest positions? How well known is the brand outside its own website? Those questions belong to the established SEO discipline, and they remain central when the discovery interface changes.
What are teams implementing under the name GEO?
The implementations I see share a shape: markup, formatting, and volume. Schema to make pages machine-readable. Answer-style formatting to make passages extractable. More pages on more questions to increase surface area. None of this is nonsense — some of it is hygiene. What I notice is what the list never contains: any account of how an answer is actually assembled, or why one source is trusted over another. The checklist treats the engine as a stricter version of Google. The teams are optimizing for a ranking the engine is not running.
Why does that work miss how an answer is assembled?
I would resist selling AI visibility as a replacement for SEO. I see it as a more complicated continuation of the same underlying work: helping the right customer discover, understand, and choose the right business.
My working sequence favors foundations before amplification. If pages cannot be crawled effectively, if important information is missing, or if the site does not make its offer intelligible, it makes little sense to promise that extra PR activity will solve the underlying problem. Once the technical structure is sound, I move into better explanations, original content, relevant entities, and credible external mentions. The work is integrated: the website, search presence, editorial substance, and off-site reputation support one another.
An answer is assembled from passages the engine selects as trustworthy for each sub-question it derived from the user’s query. The selection is the game. The checklist items I see deployed — schema, formatting, surface area — address extractability, which is whether a passage can be used, not whether it will be chosen. I have watched perfectly marked-up pages lose to plain pages that said something specific, verifiable, and distinct. The team optimized the envelope. The selection was about the letter.
Why do engine-by-engine tactics become noise?
Read more: more about Miklos Roth.
The false explanation is tooling: the company needs a GEO platform, engine-specific playbooks, a monitoring dashboard. I have watched teams build separate tactics for ChatGPT, Perplexity, and AI Overviews, and the tactic lists diverge until the team is maintaining three superstitions. The engines genuinely differ — the same page can be cited in one and ignored in another — but the team’s response treats the difference as a targeting problem. In the cases I watch, it is a signal problem wearing a targeting costume.
What pattern separates a citable passage from a campaign page?
A campaign page is written to persuade a visitor who has already arrived. A citable passage is written so that an engine assembling an answer for a stranger can lift it, stand behind it, and be right. The two look similar in a CMS and behave completely differently in retrieval. I have read pages from the same company that show both: the polished campaign page, never cited, and the unglamorous specific page, cited constantly. The company usually cannot tell me which is which without checking.
Third-party estimates are useful for direction, not direct access to a platform’s private recommendation logs. I am cautious about claims that traffic from AI tools always converts better. I have seen examples that seem to involve high-intent visitors, but an anecdote is not a population-wide conversion benchmark. For an actual client, I would examine attributable visits where measurable, inquiry quality, completed transactions, and the time or spending required to produce them. If the company shares sales and margin information, return on investment becomes a more useful compass than the number of times a monitoring service reports a brand mention.
A conventional search results page may show many links, and a business in fifth or sixth place can still receive visitors. An AI answer may mention a much smaller set of providers. That is an important commercial distinction, although the number of recommendations depends on the interface and query.
Miklos Roth
What question is worth asking instead of “are we GEO-ready”?
“Are we GEO-ready?” is a checklist question, and it produces checklist answers: yes, the schema is deployed, the tool score is up. I have watched that question get answered affirmatively in companies whose citation share was falling. The question that matters is harder to put on a project plan, because it is about what the engines can truthfully say with the company’s material — and it is the question these programs were built to avoid.
The question is not whether every channel has been included in a presentation; it is whether the combination is moving the business toward a measurable goal.
PR excerpt (80–120 words, Miklos’s own words): When a company asks me for an AI visibility strategy, I do not begin with a fashionable acronym. I begin with the business: what the site is doing in conventional search, which pages bring inquiries, how well known the brand is beyond its own website. My working sequence favors foundations before amplification, and I would resist selling AI visibility as a replacement for SEO. It is a more complicated continuation of the same underlying work — helping the right customer discover, understand, and choose the right business.
