“The AI Answers It, So Nobody’s Coming To The Site”
That line landed in a meeting where we were laying next year’s numbers side by side, and it landed hard, because nobody in the room could think of a reason it was wrong.
I certainly couldn’t. I sat there and let it stand.
It’s not a silly claim either — the summary really does sit above the links now, and has become a word people say out loud in planning meetings rather than a thing SEO people mutter about.
But accept it whole and you’ve also just argued that measuring whether the models say your name is a waste of time. That’s a large thing to concede on the strength of one meeting-room sentence.
So it’s worth knowing that somebody went and asked shoppers what they actually do after they read one of those answers.
Do People Read The AI Answer And Just Buy?
They don’t. Asked what they check before deciding on a purchase, 1,008 US consumers came out at an average of 2.4 platforms.
And 49% said they read the AI summary and sometimes click through from it anyway.
Platform here means the places a person goes for information — search, social, review sites, marketplaces. (Nobody publishes the breakdown of which ones, so I can’t tell you whether the 2.4 is mostly reviews or mostly Amazon.)
What the number does establish is that the checking doesn’t stop in one place.
Who Counted, And How They Asked
Fractl, a US content marketing firm, ran the survey in 2026 across 1,008 consumers and 150 marketers (frac.tl).
The design is a year-over-year comparison: the same questions put in 2025 and again in 2026, so the movement between the two waves is the interesting part rather than any single figure.
According to that same wave, 70% said they’re using AI in search more than they were a year ago.
So usage is climbing and the decision still isn’t being handed over. Both of those live inside one survey, which is convenient, because it means the two numbers were collected off the same panel and you don’t have to argue about whether they’re comparable.
What 2.4 Platforms Means When You’re The One Reporting The Numbers
If shoppers keep moving after the AI names you, the mention isn’t the end of the trip. It’s the door.
Somebody who saw your name in an answer then searches for the company, opens the site directly, reads a couple of reviews. That’s the 2.4 in motion.
None of which arrives in your analytics as AI traffic, because the last link they touched was a branded search or a direct visit. The trail exists; it just files itself under somebody else’s heading, and the heading it files itself under is one you’ve probably been reading as “people who already knew us”.
It’s like counting the till and never counting the door. Of course the two numbers don’t match.
Judge the AI channel on its own attributed traffic alone and you’ll write off exactly the part that worked.
Trust In AI Fell. The Number Of Places People Check Didn’t.
The share saying AI is more useful than traditional search dropped from 82% to 54% between the two waves, and the share calling it not useful went from 3% to 17%.
Fewer people, then, are impressed. Meanwhile 40% said their trust in a favourite brand goes down when that brand leans heavily on AI in its marketing — double the 20% who said so in 2025.
If trust is falling, why isn’t the checking falling with it? You’d expect a disappointed shopper to give up on the whole apparatus and go back to whatever they did before.
Maybe the opposite is what’s happening: it’s precisely because they don’t trust the answer that they go and check it somewhere else. Distrust doesn’t remove the AI from the journey. It adds stops after it.
Which puts what the models say about you in an awkward spot. Being described wrongly and never finding out about it is the expensive version of this, and 14% of the marketers surveyed said AI errors had already touched real customer relationships or revenue.
Four Things To Say Before You Put 2.4 On A Slide
Fractl sells content marketing. “You should invest in visibility in AI search” is the conclusion that sells the service, so read the direction of travel with that in mind.
The sample is US consumers and US marketers. Whether Japanese, German or Spanish-speaking shoppers also land on 2.4 isn’t something this survey can tell you.
It’s self-reported, not behavioural — a panel answering questions about what they do, not a log of what they did. The 2.4 is the number of places people say they went.
And the publisher doesn’t state the study’s limitations anywhere. The only methodological claim on the page is the year-over-year design.
I’ve taken to putting those four down before the figure rather than after it. The number survives the meeting better that way.
Read AI Mentions Next To Branded Search And Direct Visits
Put three lines on one screen and line them up in time: how your name shows up in AI answers, branded searches that week, direct visits that week.
Branded search is people typing your company or product name in as the query — only somebody who’s already heard the name can type it, which is why it moves late and reads like an echo of something that happened earlier.
I’d keep the second thing in my pocket rather than on the slide. When someone says the clicks aren’t coming so it doesn’t matter, 2.4 and 49% are enough to buy back the ten minutes it takes to explain why the two aren’t the same claim.
Losing that argument once was plenty. Try adding one line to the next monthly deck and see whether the shape of the conversation changes.
Sources
- Fractl, “AI Search Consumer Trust Study: Brand Visibility Strategies for 2026”, 2026, frac.tl (survey of 1,008 US consumers and 150 marketers, designed as a year-over-year comparison repeating the 2025 questions. Consumers check an average of 2.4 platforms before a purchase decision; 49% read the AI summary and sometimes click through; 70% report using AI in search more than the previous year. “AI is more useful than traditional search” fell 82% → 54% while “not useful” rose 3% → 17%; 40% say trust in a favourite brand falls when it leans heavily on AI in marketing, against 20% in 2025; 14% of marketers said AI errors had affected real customer relationships or revenue. Self-reported survey data, US only; the breakdown of which platforms make up the 2.4 is not published; Fractl sells content marketing; the publisher states no limitations beyond the year-over-year design.)
