The new front page is the name an AI recommends

Something has quietly flipped in how people find a supplier. Type a question into ChatGPT, Gemini or Google's AI Overview and most of the time you get a written answer, a couple of names, and no need to click through ten blue links. The prize is no longer topping page one. It is being one of the names the answer mentions.

The click is disappearing

According to Semrush, 93% of searches inside Google's AI Mode now end without a single click, and Google's own AI Overviews close out roughly 83% of the searches they appear in. Bain found that even a plain Google search ends without a click 60% of the time. Add it up and a huge share of the research a buyer does about your category never touches your website at all.

For years the job was to rank. Now a large part of that job happens somewhere you cannot see: inside a model that reads the web once, forms a view, and repeats that view to thousands of people who never visit a single page. The website is still there. It has simply stopped being the first thing a buyer meets.

This is not a niche shift confined to consumer search. Forrester's 2025 research found B2B buyers moving more slowly onto AI tools than consumers, but the traffic that does arrive from those tools is growing roughly 40% month on month, off a small base. Small today, compounding fast.

Ranking and being recommended are not the same job

Ranking rewards a page that answers one search well and earns links. Being recommended by an AI rewards a name the model has learned to associate with a category, a problem, or a claim. It is closer to how a good supplier gets mentioned in a room they are not in, because the story about them is clear, consistent, and easy to repeat.

That takes a different kind of work. A model cannot click through to check your nuance, weigh up your homepage design, or read between the lines of a clever headline. It works from whatever clear, well structured, repeatable facts about you it can find, on your own site and everywhere else you are mentioned: review sites, directories, forums, other people's articles.

A business can rank well and still be invisible to a model, if everything it says about itself is vague. And a much smaller business can earn a mention ahead of a bigger rival, simply because its claims are specific enough for a model to quote with confidence.

What actually earns a mention

Four things move the needle more than any technical trick:

• One clear, canonical page per topic, so the model is not choosing between five vague versions of your point
• Plain factual claims a model can quote directly, rather than adjectives that need interpreting
• Real evidence, a number, a named source, a specific result, rather than a generic promise
• Consistent mentions across other sites, reviews and directories, so the model sees the same story more than once

Structured data and FAQ markup help a model parse a page quickly, but they are a formatting aid, not a substitute for having something specific and true to say. Get the facts right first. The markup is decoration on top.

How to tell whether it is working

Ask the models directly. Put your category's most common buyer questions into ChatGPT, Gemini and Perplexity and see whether your name comes up, and what it says about you when it does. Do this every month, keep a simple log, and track the change over time. Treat it the same way a good salesperson tracks what prospects are saying about them in the market.

Then watch two numbers on your own site: branded search volume, meaning whether more people search your name after hearing it from an AI, and the small but growing slice of referral traffic that arrives from AI tools directly. Neither number will be large yet. Both should be moving in one direction.

What this means for the next twelve months

Treat every page, case study and profile as a fact an AI model might repeat word for word, because increasingly it will. Write the plain sentence you want quoted, put a real number behind every claim, and keep one page per topic instead of five overlapping ones that dilute the story a model would otherwise repeat with confidence.

None of this replaces the basics. A page still needs to rank, load fast and convert a visitor who does click through. But the buyers who never click are no longer a rounding error. They are a growing share of the room, and right now most companies are not in the conversation the model is having about them.

You cannot buy your way into an AI's answer. You earn it the same way you earn a good reputation: by being clear, specific and consistent everywhere the model is looking.