Why are buyers asking AI before they ask you?

Summary

Buyers use an assistant to build a shortlist because it is faster than opening five tabs, and the research step now ends before you can see it. The commercial problem is not ranking lower — it is absence, which leaves no trace in your analytics because there is no impression to log. AgencyAnalytics' 2026 Benchmarks Report of 494 agency professionals found 66% reporting client demand for answer engine optimisation, and Ahrefs found in March 2026 that only 38% of pages cited in Google's AI Overviews also ranked top ten, down from 76% in July 2025. Most of what gets cited is still controllable: Yext put it at 86% in October 2025.

Published

Because it is faster, and because it works well enough. A buyer with a shortlist to build used to open five tabs and read five sets of marketing copy. Now they can ask one question and get a comparison written for them in a paragraph, with three or four companies named. The research step did not disappear — it moved somewhere you cannot see, and it now ends with a shorter list than it used to.

The commercial problem this creates is not that you rank lower. It is that you are not in the room. A page ranked eleventh still exists on a results screen. A company left out of a generated answer is simply absent, and the buyer has no way of knowing there was an omission.

Is this actually happening, or is it a vendor story?

The honest answer is that the direct evidence is thinner than the marketing around it, so it is worth being precise about what is measured and what is inferred.

What is measured: AgencyAnalytics' 2026 Benchmarks Report, based on 494 agency professionals, found 66% reporting that clients are asking about answer engine optimisation. That is a demand signal from agencies, not proof of buyer behaviour — it tells you the question is being asked in the market, not how many buyers use an assistant to build a shortlist.

Also measured: Ahrefs found in March 2026 that 38% of pages cited in Google's AI Overviews also ranked in the top ten for the same query, down from 76% in July 2025. That is a statement about a mechanism rather than about behaviour, and it is the more useful of the two, because it says the surface your buyers are now reading is not simply a reformatted version of the rankings you already hold.

What is inferred, and should be labelled as such: how many of your buyers do this. Nobody publishes that per industry, and any agency quoting a specific percentage for your category is estimating. It is testable in your own market, which is the point of the free check.

What changes when an answer replaces a results page?

Three things, and each one has a different consequence for a supplier.

Ten blue links versus one generated answer
Results page Generated answer
How many suppliers appear Ten or more, plus paid slots Typically a handful, sometimes one
What decides inclusion Ranking for the query Whether a passage was retrieved and could be attributed
What the buyer sees of you Your title and description, written by you A summary of you, written by the model from whatever it retrieved
What absence looks like Position 11 — visible if they scroll Nothing at all — invisible, and unremarked

The third row is the one that tends to surprise people. In a generated answer, the description of your company is assembled from sources rather than copied from your meta description. If the sources disagree with each other, the summary hedges. If they are thin, the summary is thin. You do not control the sentence, but you do control most of what it is written from.

Why does being absent from an AI answer cost more than ranking low?

Because a shortlist is a filter that runs once. A buyer who asks an assistant for four suppliers in a category, gets four names, and starts contacting them has finished the discovery phase. There is no second pass in which the omitted companies get considered. Ranking eleventh is a bad position in a process the buyer is still running. Being unnamed is exclusion from a process that has already closed.

This is also why the problem is hard to notice from your own analytics. Nothing appears in your traffic reports when an engine declines to mention you, because there is no impression to log and no click that did not happen. The absence has no telemetry. The only way to see it is to go and ask the question yourself.

Does this mean the buyer never visits my website?

Often the opposite, and this is the part that makes the work worth doing rather than merely defensive. Being named in an answer does not end the buying process for a considered industrial or technical purchase — it starts it. The buyer takes the three names and goes and looks at them.

So the site still has to do its job. What has changed is the order: the site is now read after the shortlist rather than as the means of getting on it, and a separate mechanism decides whether you are on the list at all. Two jobs, not one.

Which parts of this can I actually control?

More than the framing usually suggests. Yext analysed 6.8 million citations across 1.6 million AI answers in October 2025 and found 86% came from sources a brand can manage or influence — 44% first-party websites and 42% listings and profiles, with the remainder from reviews, social, news and forums. That is a controllable majority.

Ranked by how much control you have and how quickly it moves:

  1. Your own pages. Entirely yours, and the fastest thing to change. Whether a crawler can read them at all comes first, before any editing.
  2. Your listings and profiles. Yours to edit, but scattered across surfaces you probably have not looked at in two years. Consistency matters more than volume here, because engines cross-check.
  3. Third-party mentions. Influence rather than control — industry publications, association pages, partner sites. Slow, and worth knowing about rather than chasing.
  4. The model's phrasing. Not controllable at all. Anyone selling you control over this is selling something they do not have.

What is worth doing about your own situation

Before spending anything, establish empirically whether the engines your buyers use currently name you, which sources they cite when they name a competitor instead, and whether your pages are even readable to a crawler that does not run JavaScript. That is a measurement, not a strategy, and it costs nothing but the asking.

The service that follows from it, if it turns out you need one, is getting found when buyers ask AI.

Run a free AI visibility check

Frequently asked questions

Do B2B buyers really use AI assistants to choose suppliers?

The direct evidence is thinner than the marketing around it, so it is worth separating the two. AgencyAnalytics' 2026 Benchmarks Report of 494 agency professionals found 66% reporting that clients are asking about answer engine optimisation, which is a demand signal from agencies rather than a measurement of buyer behaviour. How many of your buyers do it is not published for any specific industry and is testable in your own market.

Why is being absent from an AI answer worse than ranking low?

Because a shortlist is a filter that runs once. A page ranked eleventh is still visible to a buyer who scrolls, and the process is still open. A company left out of a generated answer is not in the process at all, and the buyer has no way of knowing there was an omission. There is no second pass in which omitted suppliers get reconsidered.

Will I see this in my analytics?

No. When an engine declines to name you there is no impression to log and no click that failed to happen, so the absence has no telemetry. The only way to observe it is to ask the questions yourself across each engine and record who was named instead.

How much of what AI cites can a company actually control?

Most of it. Yext analysed 6.8 million citations across 1.6 million AI answers in October 2025 and found 86% came from sources a brand can manage or influence — 44% first-party websites and 42% listings and profiles. What is not controllable is the model's phrasing, and anyone selling control over that is selling something they do not have.

Find out what the engines currently say about you

Send me your company name and website. I run a set of buyer questions across ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews, and send back a recorded walkthrough of what came out: where you appeared, where you did not, who was named instead, and the two or three structural reasons why. No charge, no obligation, and you keep the findings whether or not you decide to hire me.

I run these myself, so there is a queue. Expect a few working days rather than an instant report.