Research

The AI Mention Effect

Two million AI conversations were analyzed to answer one question: does being named in an AI answer actually send you customers? It does. And more than 97% of those visits arrive with no record of where they came from.

By

TL;DR

  • Profound analyzed more than 2 million U.S. AI conversations from January to June 2026, isolating cases where a brand appeared in the AI’s answer but not in the user’s prompt.
  • Those mentions lifted seven-day visit rates by 2.07 to 3.21 percentage points. Gemini showed the largest multiple at roughly 2.5x baseline.
  • Only 20.5% of the visits happened within an hour and 42% within a day. The majority came later in the week.
  • More than 97% arrived with no UTM, so analytics filed them as direct or organic traffic.
  • If mentions drive visits you cannot attribute, the mention itself becomes the metric you have to measure.
Bar chart comparing baseline and post-mention seven-day visit rates across Gemini, Google AI Overviews, and ChatGPT, with a callout that 97 percent of visits carry no UTM

The Question Nobody Had Good Data On

Everyone selling AEO services makes the same promise: get mentioned by AI and you will get business. Until recently, nobody could show the second half of that sentence.

The problem is not that AI mentions are hard to count. It is that the visit they produce shows up somewhere else entirely, days later, with nothing linking it back. You can see the mention. You can see the visit. You cannot see the line between them.

A study published by Profound is the first serious attempt to draw that line, and the design is the interesting part.

How the Study Worked

Profound used a double opt-in privacy panel that captured both AI conversations and the browsing that followed, covering more than 2 million U.S. conversations across ChatGPT, Gemini, and Google AI Overviews between January and June 2026.

The critical definition is what counts as exposure. In their words, “an AI-exposure is a brand mention that does not appear in the user’s prompt but does appear in the AI response.” If you ask ChatGPT about a company by name and then visit its site, that proves nothing. This study only counted the cases where the AI introduced the brand.

To establish what would have happened anyway, they compared the seven days after each exposure against three preceding seven-day placebo windows for the same users, then filtered out anyone who had already searched for or visited the brand in the prior week. Generic domains were excluded. Confidence intervals came from a user-clustered bootstrap with 2,000 replicates.

Two limitations are worth stating up front. The panel is U.S. only, and it covers ChatGPT, Gemini, and AI Overviews. Claude and Perplexity are not in the dataset, so nothing here tells you how mentions behave on those platforms.

A Mention Roughly Doubles the Odds of a Visit

Across all three platforms, users who saw a brand mentioned were meaningfully more likely to visit that brand’s site in the following week.

Platform Baseline 7-day visit rate After a mention Lift Multiple
Gemini 2.21% 5.42% +3.21 pp ~2.5x
Google AI Overviews 4.83% 7.79% +2.96 pp ~1.6x
ChatGPT 4.33% 6.39% +2.07 pp ~1.5x

Notice that Gemini has the largest multiple but the lowest starting point, while AI Overviews produces the highest absolute post-mention rate. Both readings are useful, and they point in slightly different directions depending on whether you care about relative gain or total visits.

Read the lift in absolute terms and it stays honest: roughly three percentage points means that for every 100 people who see your brand in an AI answer, about three more visit your site that week than otherwise would have. Small per person. The size of the outcome depends entirely on how often you get mentioned, which is the number most businesses have never measured.

The Visit Does Not Happen on the Click

This is the finding that breaks conventional measurement. Only 20.5% of downstream visits happened within an hour of the conversation, and 42% within 24 hours. The majority arrived spread across the rest of the week.

Speed varied by platform. Google AI Overviews was fastest, with 45.7% of visits landing same day, which makes sense given that the user is already mid-search. Gemini was slowest at 30.0% same day.

Profound’s framing is that AI visibility currently behaves like “a billboard: its impact shows up in what people do hours and days later.” That is the right analogy, and it has a direct consequence. If your attribution window is the session, or even the day, you are looking at less than half the effect.

Your Category Changes the Size of the Effect

The lift was not uniform across industries, and the spread is wide enough to matter when you decide where to spend effort.

  • Financial services. +3.7 to +5.7 percentage points across platforms, the strongest category in the study.
  • Retail. +2.6 to +5.6 points, with the widest platform-to-platform variation.
  • Software. +3.0 to +4.0 points, the most consistent performer.
  • Telecom. +1.5 to +3.0 points, the weakest of the four.

The pattern tracks how much research a purchase warrants. Categories where people weigh options and want a name they can verify show the biggest response to an AI recommendation. Categories where the choice is largely predetermined show the least.

Because the ranking of platforms also shifts by industry, there is no universal answer to which engine deserves your attention. It depends on your category, which means it has to be measured rather than assumed.

The 97% You Cannot See

Now the uncomfortable part. Profound found that more than 97% of these post-mention brand site visits arrived with no UTM parameter at all.

ChatGPT is the only platform that has been visibly improving here, tagging roughly 1.0% of post-exposure visits before May 2026, then 1.79% in May and 2.47% in June. That is real movement, and it is still a rounding error against the total.

Everything untagged lands in your analytics as direct or organic. Not mislabeled AI traffic. No label at all.

If 97% of the traffic is invisible, then referral traffic is not a measurement of your AI visibility. It is a measurement of how aggressively the platforms happen to be tagging their outbound links this quarter.

This reframes something we wrote about earlier. When we looked at the AI-referred traffic we could actually see, it engaged longer and converted at close to three times the rate of standard referral traffic. Put the two findings together: the visible sliver is unusually high quality, and it represents under 3% of the real volume.

What This Changes About Measurement

Three practical adjustments follow from the data.

  • Measure mentions, not clicks. The mention is the event that causes the visit, and it is the only part of the chain you can observe reliably. Mention rate, sentiment, and share of voice are now leading indicators rather than vanity metrics. Our guide to tracking AI citations covers the five that matter.
  • Widen the window. A seven-day view is the minimum that captures the effect. Watch direct traffic and branded search volume alongside your mention data, because that is where the majority of AI-influenced visits are actually landing.
  • Prioritize by category, then platform. Given how much lift varies by industry, the platform worth your effort is an empirical question about your own category, not a matter of market share.

The through line is simple. AI assistants are influencing purchase decisions in a way that leaves almost no forensic trace in your analytics. The only reliable place to look is upstream, at whether you are in the answer in the first place.

If you have not established that baseline, run your free AI visibility audit. It tests ChatGPT, Claude, Gemini, and Perplexity with structured prompts, scores how often and how accurately you appear, and emails the report within a few hours. No account, no credit card.

Sources

Frequently asked questions

Common questions about AI mentions and traffic

The AI mention effect is the measurable increase in website visits that follows a brand being named in an AI assistant’s answer. In a study of more than 2 million U.S. AI conversations between January and June 2026, Profound found that users who saw a brand mentioned visited that brand’s site at 1.5 to 2.5 times the baseline rate over the following seven days. Only mentions the user did not ask for by name were counted, which separates the influence of the mention from existing brand intent.

Yes. Only 20.5% of downstream visits happened within an hour of the conversation and 42% within 24 hours, which means most visits came days later through a separate search or a direct visit rather than a click on a citation. A mention behaves less like a search result and more like a billboard. The effect shows up in what people do hours and days afterward, whether or not a clickable link was present.

Because most AI-influenced visits carry no tracking parameters. More than 97% of post-mention brand site visits arrived with no UTM. ChatGPT tagged roughly 1.0% of post-exposure visits before May 2026, rising to 1.79% in May and 2.47% in June. Everything untagged lands in Analytics as direct or organic traffic, so a referral report understates AI-driven visits by a very wide margin.

It depends on whether you measure the multiple or the absolute lift. Gemini showed the largest multiple, taking seven-day visit rates from 2.21% to 5.42%, roughly 2.5 times baseline. Google AI Overviews produced the highest post-mention rate at 7.79%, up from a 4.83% baseline. ChatGPT rose from 4.33% to 6.39%. Lift also varied by industry, from 1.5 percentage points in telecom to 5.7 points in financial services, so the right platform to prioritize is a question about your own category.

No credit card required

Find out how often AI mentions you.

Free audit across ChatGPT, Claude, Gemini, and Perplexity. Scored across five dimensions. Results emailed within a few hours.

Run your free AI visibility audit