Why AI is changing how buyers research firms
For years, when someone needed a law firm, an accountant, or an agency, the process was predictable. They search Google, open tabs, scan websites, and build their shortlist right in the open. You could see the journey unfold, and it usually led straight to your front door.
That sequence has changed. Now, the shortlist forms inside an AI tool—often before a buyer even visits a website, and usually without the firms ever knowing they were in the running. I want to show you what’s changed, what the data reveals, and why ranking on Google is no longer the same as being found.
The decision now happens before the first click
The part of the buying journey that used to play out on your website now happens out of sight. Prospects asking ChatGPT, Perplexity, or Gemini who the best firms are in their field. The tool responds with names, a short description, and a reason to trust each one. By the time someone lands on your site, the shortlist is set and the questions are already shaped.
This is the shift that matters. Your website used to be where buyers made their decision. Now, it’s where they confirm a choice an AI tool has already shaped. If the tool leaves you out or gets your story wrong, your website never even gets a shot.
What the evidence shows
This is not a forecast. It is already how buyers behave.
of B2B buyers used AI tools somewhere in their 2025 purchase journey, according to research compiled by Deep Marketing.
of B2B software buyers now begin their research in an AI chatbot rather than a search engine, up from 29% a year earlier, per TechEdge AI.
of UK B2B buyers use AI tools before they make first contact with a supplier, rising to 75% among 25 to 40-year-olds, in Whitehat SEO’s 2026 UK research. The same study found more than 60% of UK SMEs have used AI to research accountancy or business advisory services.
The pattern holds across sources: the earliest, most decisive stage of buying now happens inside tools most firms have never tested themselves against. Altair Media’s analysis puts it bluntly. The shortlist and the cross-checking that follows now carry about 70% of the decision weight, with almost no direct contact with the firms involved.
What stages one and two now look like
The first two stages of a B2B decision used to be a search and a skim. Now, they look and feel like a conversation.
Stage one is the shortlist.
A buyer describes their problem to an AI tool and asks who can solve it. The tool returns a set of names. You are either in that set or you are not, and the buyer rarely sees the firms that were left out. This is the moment a firm becomes invisible, without ever hearing a no.
Stage two is the cross-check.
The buyer, now holding a shortlist, starts asking follow-up questions. Which of these firms has real experience in my sector? What do people say about them? Can this one handle the size of the job I have? The tool pulls from whatever it can find and trust: your website, third-party mentions, reviews, and published work. If your evidence is thin or unclear, you lose here, quietly, to a competitor who looks better supported.
Picture two accountancy firms in the same city, both competent, both with tidy websites. A buyer asks an AI tool for the best mid-market accountants nearby. The first appears by name, described accurately, with a line about the sectors it specialises in and a nod to a published client result. The second is not mentioned at all, because nothing on its site tells the tool what kind of work it does or who it does it for. Neither firm sees the question. Only one makes the shortlist. The difference isn’t the quality of service. It’s whether the tool found enough clear, trustworthy evidence to act on.
By the time a prospect lands on your site, both of those stages are usually done. Your website isn’t making the case anymore. It’s being checked against a case the buyer already heard.
Why ranking on Google is no longer the whole picture
Search engine optimisation still answers one question well: can your pages rank when someone types a query into Google? That question still matters. But it’s no longer the whole picture, because more and more buyers aren’t typing a query into Google at all.
The two jobs are different.
Shows a list.
Google ranks pages and shows the buyer a list to choose from.
Hands an answer.
An AI tool reads the web, forms a view, and hands the buyer an answer with the choosing already done.
You can rank well on Google and still be absent from the AI answer, because the tool weighs different signals: whether it can find you, whether it understands what you do, whether it trusts you enough to recommend you rather than merely mention you, whether it can use your site, and whether your content is structured clearly enough to quote.
There is a measurable upside to getting this right. Whitehat SEO’s research found that visitors arriving from a ChatGPT referral converted at 15.9%, against 1.76% for Google organic search.
What an AI tool needs to recommend you
An AI tool recommends a firm when it can do five things easily.
-
01
It needs to find you when someone describes your kind of work.
-
02
It needs to understand what you do without guessing.
-
03
It needs enough trustworthy signal to recommend you rather than merely list you.
-
04
It needs to be able to use your site if a buyer asks it to act.
-
05
And it needs your content structured clearly enough to quote.
Most of what makes a firm easy to recommend is unglamorous. Clear, specific descriptions of what you do and who you do it for, in plain language, on the pages that matter. Third-party evidence the tool can find: named case studies with real outcomes, mentions in places it already trusts, reviews, published work in your field. Content laid out so a machine can read it, with proper headings, named authors, dates, and claims it can lift and attribute. Technical basics that let an automated visitor move through the site and complete an action without breaking.
None of this is exotic. It is the same instinct that has always separated a firm that is easy to choose from one that is merely good at the job. What has changed is the reader. The first reader of your evidence is now often a machine, deciding in seconds whether you are worth putting in front of a person.
What this means for an established firm
The firms most exposed to this shift are not the ones ignoring AI. They are established firms that built a solid web presence for the search era and assume it carries over. Often it does not. A site can be well designed, well written and well ranked, and still confuse an AI tool about what the firm actually does.
The evidence on the supply side is striking.
A 2026 study of 110 UK law firms by AireStream found that one in six were effectively invisible to AI assistants, and half scored below 60 out of 100 on AI discoverability. These are not failing firms. They are competent businesses whose evidence is thin, buried or unclear in exactly the ways an AI tool penalises.
For a firm doing several million a year in revenue, the risk is specific. You can keep investing in marketing, content and the website, and still not know whether AI tools find you, describe you correctly, or recommend a competitor more confidently than they describe you. The spend continues. The picture of whether it is working has a blind spot in it, and the blind spot is now where the decision starts.
I spent a decade running experimentation programmes before AI rewrote the rules, for organisations including Cancer Research UK, lastminute.com and Vodafone. The discipline there was always the same: do not guess what is working, test it, measure the gap, and fix the biggest one first. That discipline transfers cleanly to this problem. The question is not “should we do more AI”. It is “what are AI tools actually saying about us, and is it accurate”.
The question worth answering first
Three questions sit underneath everything on this page, and a firm can ask them today.
-
When a buyer asks an AI tool for the best firms in our field, do we appear, or do our competitors?
-
When an AI tool describes us, does it get the description right?
-
When it decides who to recommend, does it find enough evidence to choose us?
Most firms have never checked. The tools are free, the competitors are named, and the answers are sitting there to be read. The reason it rarely gets done is that reading those answers properly, across several tools, against named competitors, and turning the result into a list of things to fix, is a piece of work in its own right.
If you want to see exactly where your firm stands, that is what the AI Visibility Scorecard does: it tests what ChatGPT, Perplexity and Gemini say about your business against three competitors you name, reviews the pages that matter most, and returns a prioritised list of what to fix first.
AI Visibility Scorecard