Mostly, no. An AI assistant answers the question on its own screen, names a few sources, and the person moves on without clicking. Some traffic does come through, and it is growing from a small base, but anyone expecting AI citations to replace search clicks one for one is reading the data backwards. The point of being cited by ChatGPT or Google's AI answers is not the referral visit. It is that your version of the facts is the one the assistant repeats to everyone who asks, whether they click or not. Here is what the numbers show, where the traffic actually goes, and how to think about AI visibility once you accept that the click is not the prize.
How many people are even using AI to find things?
Enough that this is no longer a question about early adopters. The Pew Research Center's Americans and AI 2026 report, released June 17, 2026 and based on a survey of 5,119 US adults fielded February 17 to 23, found that 49% of US adults now use AI chatbots, up from about a third in 2024. Roughly one in four use them daily.
The Reuters Institute's 2026 Digital News Report, published June 16 and drawn from 97,520 respondents across 48 markets, found that weekly use of AI chatbots for news rose from 7% to 10% year on year. That number is smaller because it is specific to news, but the direction is the same across every measure: more people, every year, asking an AI tool first. OpenAI itself has since put a number on just how much that volume has grown: ChatGPT's search-like query volume versus Google's is now large enough that the company is willing to compare it in public, however unverified the exact figure is.
So the audience is real and it is growing. The question for any brand is what happens after the assistant answers.
Source: Pew Research Center analysis of the browsing data of 900 US adults, published July 22, 2025, linked in the article.
What happens to clicks when an AI answer is on the screen?
They fall by about half. This is the finding that matters most, and it comes from observed behavior, not a survey. Pew analyzed the actual browsing data of 900 US adults and published the results on July 22, 2025. When Google showed an AI summary at the top of the results, users clicked a traditional search result link in just 8% of visits. When no AI summary appeared, they clicked nearly twice as often, in 15% of visits.
The link inside the AI summary fared worse. People clicked a source link in the summary itself in only 1% of visits to pages that had one. And the sessions ended sooner: 26% of visits to a page with an AI summary were the last thing the person did before leaving, compared with 16% of visits without one.
That is the mechanism. The AI answer satisfies the question, so the reason to click is gone. This is the same dynamic SEO and AI citations now pull against each other, where ranking first no longer guarantees the visit it used to.
But don't people say they click through from chatbots?
They say they do, and the gap between what people say and what they do is the whole story. The Reuters Institute found that 42% of people who use AI chatbots for news say they always or often click through to the original source. Taken alone, that sounds reassuring.
Set it next to Pew's 8%, though, and the problem is clear. One number is self-reported intention; the other is logged behavior. People overstate how often they check sources, the same way they overstate how often they floss. The two studies measure slightly different things, Google's search summaries versus standalone chatbots, but they point the same way: the click people believe they make is far more common than the click they actually make.
There is a trust angle underneath this too. Reuters found that global trust in AI chatbot answers sits at 20%, against 37% trust in news overall. People lean on these tools while only half-believing them, which is exactly the behavior that makes clicking through feel optional. For why that mistrust does not stop the citation from doing its work, it helps to separate the two things a citation does.
So is a citation worthless if nobody clicks?
No, and this is where most brands get the logic backwards. A citation does two jobs. The first is to send a visitor, and at that job, as the data shows, it is weak. The second is to decide whose facts the assistant repeats, and at that job it is the only thing that matters.
When someone asks ChatGPT which companies do a certain thing, or whether a particular brand is reputable, the answer is assembled from sources the model trusts. If your company is described accurately in those sources, the assistant repeats the accurate version to every person who asks, none of whom you ever see in an analytics dashboard. If your company is absent or described wrong, that is what gets repeated instead. The visit was never the point. Being the source was. This is the core argument behind earned media's share of AI citations: the coverage an assistant leans on is doing the work whether or not it ever produces a click.
How should you measure AI visibility if the click isn't the point?
Not with your referral traffic report, or at least not with that alone. Referral traffic from AI tools is real but small, and judging the whole effort by it is like judging a billboard by how many people phone the number on it. Most of the value is the impression, not the call.
The measures that actually track AI visibility are different: whether the assistant names you when asked about your category, whether it describes you correctly, and which sources it cites when it does. How to measure generative engine optimization walks through the practical version of this, testing the real prompts your buyers use and recording what comes back. Treat the occasional referral click as a bonus, not the scoreboard.
What should you actually do about it?
Work on the source material, not the click. Three things move the needle.
First, get the facts about your company stated plainly and consistently everywhere an assistant might read them: your own site, your press kit, your directory listings, and the coverage written about you. An assistant resolves conflicts by guessing, and a consistent public record gives it less to guess about.
Second, earn coverage in places that get cited. A mention in a publication an AI system already trusts carries more weight than anything you publish about yourself, which is why getting cited by ChatGPT runs through the same earned-media work as traditional PR, not through a separate set of tricks.
Third, stop optimizing for a click that the format has quietly removed. The brands adapting fastest are the ones that accepted the new shape of the page: generative engine optimization is about being the answer, not about winning a race to a link that fewer people press each year.
The direct answer, for anyone who read this far
AI chatbots send some traffic, growing slowly from a small base, but far less than their reach suggests. Pew's logged data shows clicks roughly halving when an AI answer appears, from 15% to 8%, and falling to 1% for the links inside the answer. People say they click through far more than they do. The useful conclusion is not that AI visibility doesn't matter; it is that the citation matters more than the visit. When half of adults ask an assistant first and most never click, the thing worth owning is the answer itself, which means getting your facts straight everywhere and earning coverage in the sources these tools actually read.
If you want to see how AI assistants currently describe your brand, and where their answers come from, request a quotation and we'll show you what's out there.



