For 20 years, being found online meant ranking on a page of blue links. That is breaking. More buyers now ask an assistant a question and get a single answer, with two or three named suppliers, and never see a results page at all. If the AI does not know you, or does not trust what it reads about you, you are simply not in the conversation. Here is what is changing, how the assistants actually choose who to name, and what to do about it.
The shift, in one line
Search used to send you a list to choose from. An AI answer makes the shortlist for you. That moves the battle from ‘rank on page one’ to ‘be the source the AI quotes’.
The difference matters more than it sounds. On a results page, position four still gets clicks. In an AI answer, there is no position four. There are the two or three names the assistant chose, and everyone else is invisible. The funnel has a much narrower neck, and the criteria for getting through it are different from the ones most firms have spent a decade optimising for.
The numbers behind it
In 2024 Gartner predicted traditional search volume would fall 25% by 2026 as people move to chatbots and assistants. That exact figure is now debated, and some analysts argue the drop has been softer than forecast. Fair enough. The direction is not in doubt.
ChatGPT reached around 900 million weekly users by April 2026, more than double the figure 14 months earlier. Google’s AI Overviews, the answer box above the links, now reach roughly 2 billion people a month. Perplexity handles tens of millions of queries a day and is growing far faster than traditional search. A large and growing share of the questions your buyers ask are answered by an AI before a single website is clicked.
The part that should concentrate the mind is that this is already true for business-to-business buying. A founder looking for a fractional CTO, an operations director looking for a technology review, a finance director checking a supplier, increasingly asks an assistant first and uses the answer as the shortlist. The old habit of opening ten tabs is fading.
How an assistant actually chooses who to name
This is the part most firms get wrong, because they assume it works like Google. It does not, quite.
An assistant reads a question, decides whether it needs fresh information, fetches a set of pages, and then does three things. It looks for a passage that directly answers the question in the words the question used. It judges whether the source is credible, using signals such as named people, specific facts, structured data, and whether other trusted sites mention the same source. And it prefers passages it can lift cleanly, a sentence or two it can repeat without having to untangle marketing language.
So three qualities decide whether you get named. Directness, credibility and quotability. A page that buries the answer, hedges everything and has no third-party backing loses to a plainer page that says what it does and is mentioned elsewhere.
What makes you the source an AI quotes
Three things decide it, and they compound.
Answer the question directly, high on the page, in the words a buyer would use. An AI lifts a clean sentence far more readily than it untangles marketing copy. If the question is ‘who offers fractional CTO support in Cardiff’, the page should say, in a sentence, who offers fractional CTO support in Cardiff.
Be specific and attributable. Numbers, named people, real credentials, concrete examples. Assistants favour sources that state things plainly and back them up, because that is what they can safely repeat. ‘We have deep experience’ is unquotable. ’20 years in technology leadership, former CTO and VP Engineering, Fellow of the BCS’ is a fact an assistant can use.
Earn mentions off your own site. The domains AI trusts most in any sector are usually not the supplier sites, they are LinkedIn, industry bodies, directories and the press. Being referenced there is what lifts you into the answer across different assistants. Your own site gets you considered. Other people’s sites get you chosen.
What this looks like in practice
When we tracked our own visibility across the main assistants, the pattern was clear and a little humbling. The pages that got cited were the ones with a direct answer in the first paragraph and a short FAQ at the bottom. The pages that did not get cited were the better-written ones, the ones with a narrative opening and the answer three paragraphs down. Assistants rewarded clarity over craft.
Off-site, the sources assistants pulled from most in our sector were LinkedIn, an industry body we belong to, two directories and a regional business news site. None of them were competitor websites. That told us where to spend effort, and it was not on the homepage.
The changes that moved the needle were cheap. Direct-answer openings on the service pages. FAQ blocks with the questions buyers actually ask. Schema markup. A current member profile on the industry body’s site. Nothing clever, all of it visible to a machine.
What to measure instead of rankings
Stop watching only your Google position. Track three things instead.
How often an AI mentions you for the prompts that matter, across the main assistants. Whether it cites your own pages when it does, or points somewhere else. And which assistants feature you and which do not, because they differ, and a gap on one is usually a gap in a specific kind of source.
Tools exist to monitor this, and the monthly trend tells you far more than a single keyword rank ever did. Pick five buyer questions, check them monthly, and watch the trend.
The common mistakes
Writing for the algorithm you grew up with. Keyword-stuffed pages that rank on Google often read as untrustworthy to an assistant, because they are vague and repetitive. Write for a sceptical human and the assistant follows.
Hiding the answer. Narrative openings, brand stories and scene-setting paragraphs push the quotable sentence below the fold. Put the answer first and the story after.
Ignoring the off-site work. Most firms pour effort into their own site and none into the directories, bodies and publications assistants actually trust. The split should be closer to half and half.
Measuring the wrong thing. A rankings report that says you are fourth for a keyword tells you nothing about whether an assistant names you. Measure the mentions.
The honest trade-off
This is slower to move than paid ads and harder to control than a landing page. You cannot buy your way into an AI answer, at least not yet, and you cannot fake the authority signals for long. The work is real content and real credibility, built over months. The upside is that once you are the cited source, you are hard to dislodge, because the same authority that got you there keeps you there, and a competitor cannot outbid you for it.
Where to start
Pick the five questions a good customer would ask an assistant before they buy. Check what the assistant says today, and whether you appear. Then make the pages that answer those questions the clearest, most quotable source available, and get your firm accurately listed in the two or three places the assistant already trusts in your sector. Our piece on getting your business ready for AI agents covers the technical side. An AI opportunity review is a sensible way to map all of it, and our Insights library is built exactly this way.
Frequently asked questions
What is agentic or AI search?
It is when a buyer gets a direct answer from an AI, such as ChatGPT, Perplexity or Google’s AI Overviews, rather than a list of links. The AI reads the web and names a few options, so the goal shifts to being one of those named options.
Is traditional SEO dead?
No. Classic search still drives large volumes, and much of what helps AI answers, clear content and strong authority, helps ordinary rankings too. The change is that you now write for being quoted as well as for being clicked.
How does an AI decide which businesses to name?
It looks for a direct answer to the question, judges whether the source is credible using specific facts, named people, structured data and third-party mentions, and prefers passages it can quote cleanly.
How do I get my business into AI answers?
Answer buyer questions directly on well-structured pages, be specific and credible, and earn mentions on the third-party sites that assistants already trust in your sector.
How long does it take?
Months rather than weeks. On-site changes can show within a few weeks. Off-site authority takes longer to build and longer to lose, which is the point.
How do I measure success?
Track how often assistants mention and cite you for your key prompts, and watch the monthly trend by assistant, rather than relying on a single search ranking.