GEO, Measurement
Almost nobody clicks a source inside an AI answer: about 1% of visits, per Pew. Yet visitors who arrive from AI tools convert up to 54% better, per Adobe. Both numbers are true. The KPI that reconciles them is value per visit. Here is what to measure monthly.
Is AI traffic worth it? Two numbers, both true, seem to disagree. People click a source link inside an AI answer about 1% of the time. And visitors who do arrive from AI tools convert up to 54% better than everyone else. If you are weighing ChatGPT referral traffic quality against the effort of showing up in AI answers at all, you need both numbers and the argument between them. The doom camp and the hype camp are each reading one number and ignoring the other. Here is the full picture, with sources, and what we would measure instead of sessions.
The small number first. Pew Research Center tracked the real browsing behaviour of 900 US adults. When a Google results page included an AI summary, users clicked a source link inside that summary in about 1% of visits. The wider trend points the same way: SparkToro's analysis of Similarweb clickstream data found that 68.01% of US Google searches ended without any click between January and April 2026. And the referral pie has shifted less than the headlines suggest: Cloudflare Radar referral data showed Google still sending about 88% of search referrals in July 2026, with ChatGPT at about 0.9%.
Now the big number. Adobe's Q1 2026 analysis of more than 1 trillion visits to top North American retail sites found that visitors arriving from AI assistants converted 42% better than non-AI traffic in March 2026, and generated 37% more revenue per visit. By May the gap had widened: Adobe Analytics figures reported by Digital Commerce 360 show AI-referred retail traffic up 138% year on year and converting 54% better than non-AI traffic.
So: almost nobody clicks, and the few who do are among the most valuable visitors your site gets. Hold both. The doom take ("the web is dead") reads the first number and stops. The hype take ("AI traffic will replace Google") reads the second and forgets that AI referrals are still around 1% of the pie.
A zero-click search is one where somebody asks a question, reads the answer on the results page, and leaves without visiting any website. About 68 in 100 US Google searches now end that way, per the SparkToro data above.
That sounds like extinction. For a small business it is something stranger: the visit is disappearing, but the recommendation is not. When a buyer asks an assistant for the best accountant for contractors in Leeds and the answer names three firms, one of those firms gets a phone call that no analytics tool will ever attribute. There was no click. There was still a customer.
This is why we keep banging on about mentions. Presence in the answer is the asset; the click is a bonus. We wrote about the difference between being named and being recommended in AI mentions vs recommendations, and about which mentions actually count in three ways a brand appears in an AI answer.
The Adobe numbers make sense once you think about what an AI answer does to a buyer before the click. The assistant has already summarised the options, compared them and often recommended one. Someone who clicks through after all that is not browsing. They are checking one last thing before acting. The AI answer absorbed the research phase; your site receives the decision phase.
Two honest caveats before you build a strategy on this.
First, Adobe's data is US retail. If you run a B2B service firm in Manchester, your curve will not match a North American retailer's, and nobody has published your niche's curve yet.
Second, analytics can only see sessions that arrive with an AI referrer attached. It cannot see the buyer who was given your name by ChatGPT, then typed it into Google or went straight to your site. That journey shows up as branded search or direct traffic. So the 1% click figure is a floor for AI influence, not a ceiling.
Put the numbers together and the conclusion is not "chase AI traffic" or "ignore AI traffic". It is: change the KPI (key performance indicator: the number you steer by).
Sessions were a sensible KPI when every buyer journey ran through a click. That era is ending. If total sessions fall while value per session rises, a traffic dashboard shows decline while the till shows growth. The dashboard is lying by omission.
A small business does not need AI traffic volume. It needs to be the answer when one high-intent buyer asks a question it can win. One visit from that buyer is worth more than a hundred sessions of idle skimming.
Five numbers, none of them exotic:
Track these monthly and the doom-versus-hype argument becomes irrelevant to you. You will know what AI answers are actually worth to your business, which is the only version of "is AI traffic worth it" that matters.
The two numbers are not a contradiction. They describe a funnel that moved: research happens inside the answer, decisions happen on your site. Fewer visits, better visits, and a growing share of influence that never registers as a visit at all.
Number five on the list is the one we can help with today. Get your baseline: run a free scan at whocanfindme.com. No signup. Takes about ten seconds.
When a buyer asks an AI one question, the engine quietly turns it into about ten. Seer Interactive measured 10.7 hidden sub-queries per prompt across 501 prompts on Gemini 3. Most sites answer the headline question and none of the ten. Here is how to find your gaps.
Leer másThe same question, same engine, same day can produce different answers. That is the nature of AI search, not a bug. The 3 rules we use to avoid showing you fake statistics, and how to read a visibility score without fooling yourself.
Leer más