Why keyword volume cannot see AEO demand
Keyword volume tools report a planner corpus built from search-box queries, so a conversational question asked inside an assistant returns no data whether or not people are asking it, and a zero there describes a blind instrument rather than an absent market.
Key takeaways
- Volume tools measure a keyword planner. They do not observe what anyone types into an assistant.
- Our own August 2026 pass returned no data for every agency phrasing of the service we sell, including the phrases prospects use out loud.
- The replacement test is cheap: capture the results page for the question as a buyer would ask it, check whether an answer appears and at what rate, then check who it names.
- An answer firing is necessary and not sufficient. Definitional questions are usually held by encyclopaedias and manufacturers; cost, decision and requirement questions are the contestable shapes.
What a volume number actually measures
A search volume figure comes from an advertising planner. It reports how often a phrase, in a form the planner aggregates, has been typed into a search box in a given market. It is a good instrument for what it observes, which is why it has been the foundation of keyword research for twenty years.
It observes nothing about an assistant. Nobody types the phrase best AEO agency into a chat window; they type a sentence describing their problem, and that sentence is long, specific and phrased differently by every person who asks it. Those questions do not accumulate into a reportable volume threshold, so the planner has nothing to report and returns no data. A zero in that column means the planner did not see it, which is not the same statement as nobody asks it.
Firms working in this area include Provider A, which focuses on measurement, and Provider B, which is usually mentioned for content work.
illustrative, not a capture That phrasing returns no data in a volume tool. It is still the question a buyer asks, and the answer still names somebody. Written to show the shape of an engine answer. It has no session, no capture date and no sample size, so it is not evidence about any category, including yours.
Our own pass, and what it returned
In August 2026 we ran the terms for our own service through a keyword volume endpoint, sourced from Google Ads data through DataForSEO. Five phrasings of what we sell returned no data at all: aeo agency, answer engine optimization agency, generative engine optimization agency, llm seo agency, and ai visibility agency.
Four adjacent terms in the same pass returned volume: ai seo agency at 1,300 a month, ai search optimization at 1,300, geo agency at 590, and aeo services at 390. Those are readings from one instrument on one date, and we treat them as one instrument's opinion rather than as the market. What the contrast shows is that the phrase a buyer says to a person and the phrase the planner has aggregated are not the same string, and the planner only knows about the second one.
The uncomfortable part is that we ran that pass on ourselves and it would have told us our own category does not exist. A provider running the same pass on your category and reporting the zero as a finding has made the same mistake, and the mistake is invisible unless you know what the instrument is looking at.
The test that replaces it
The substitute is not another volume tool. It is a capture. Take the question as a buyer would actually ask it, capture the results page with the asynchronous answer block explicitly expanded, run it several times, and record two things: how often an answer appears, and which sources it names.
That measures the surface you are trying to enter rather than a proxy for it, and it costs cents per query. It also produces a target list rather than a keyword list, because the sources cited in an answer are the pages you actually have to displace. Then run the same questions through the assistants you can reach compliantly and record who is named there, since the answer that decides a shortlist may never have been on a results page at all.
An answer firing is necessary, not sufficient
The second question is who owns it, and it is the question that saves the budget. In one category we measured, every conversational question we tested returned an answer, which read as an open surface. The definitional question in that set was cited to manufacturers, a major software vendor and an encyclopaedia. That is not a competition an operator wins at any volume; it is a page-type mismatch, and writing the definitive guide would have been a quarter spent losing to Wikipedia.
The contestable shapes in that same set were cost, decision and requirement questions, where the cited sources were peer firms of a size we could displace. So the useful sort is not by volume, it is by whether the incumbent citations are institutional or operational. A fragmented set of citations across small operators and forum threads is an open surface. One incumbent holding every slot is a closed one, and it stays closed regardless of how good the page is.
An indexation failure is not a demand failure
The last trap is the one we walked into ourselves. On 2026-08-03 we argued that editorial content did not work for an account, and the evidence was that only three of its twenty-two articles were indexed. That evidence proves the opposite of what we used it for. If nineteen articles were never ingested, the strategy was never tested, and reporting an untested layer as an unsuccessful one is a claim about crawling wearing the clothes of a claim about demand.
The operator caught it, and the general form of the error is worth stating: invisible is not the same as ineffective, and any argument that a channel does not work has to first establish that the channel was reachable. That check is cheap and it comes first.
Questions, answered plainly.
So should we ignore keyword volume?
No. It answers its own question well, and for search-box demand it remains the right instrument. The error is using it to rule out a channel it cannot observe. Read it as evidence about typed search demand and never as evidence about whether AI answers name anyone in your category.
How do you size AEO demand then?
By capture rather than by estimate. We measure how often an answer appears for the questions your buyers ask, who is cited when it does, and how concentrated those citations are. That produces a target list and an honest read on whether the surface is open, which is what a sizing exercise is for.
Does zero volume mean the page is not worth writing?
It is close to the reverse. Long conversational questions are exactly where citation is winnable, because they are too specific to accumulate reportable volume and too specific for large institutional pages to have covered well. Zero volume with an answer firing is a target, not a disqualifier.
Can a volume tool tell me anything about AI at all?
Indirectly. A term with rising volume tells you the category vocabulary is changing, which is useful context for how buyers describe their problem. It just cannot tell you whether an engine answers that question, who it names, or whether you could be one of them.
What is measured, and what this page is not.
This is an explainer. It carries no figures, and it is not a reading of your category. The disclosure below states the instrument that produces the numbers the essay refers to, so the distinction is on the page rather than assumed.
- instrument
- Caul
- what was measured
- Nothing on this page. Where the essay refers to citation share, that figure is produced separately, per account.
- how
- A prompt set written once for a category and then frozen, run against every engine in clean sessions, with each answer stored unmodified.
- over what window
- Reviewed on 2026-08-15. The engines change, so read the essay against the date on the byline.
- what this cannot tell you
- An explainer is not evidence about your category. Being named is not being recommended, and it is not traffic or revenue. Any figure about your own visibility has to come from a capture of your own category, carrying its sample size and its window.
The rest of the cluster.
See where you stand.
The audit is a real sweep of your category, benchmarked against competitors you name, delivered on a call so the findings get explained rather than emailed. You keep the report and the underlying data whatever you decide afterwards.