Does AI Search Send Traffic to Websites?

The question clients actually ask me is not how to get cited by an AI assistant. It is whether any of this is sending them customers. They have read the headlines, they suspect their own analytics do not reflect any of it, and they want a straight answer with numbers attached rather than a promise.

The straight answer is yes, AI answer surfaces do send visitors. The number is currently small as a share of sessions. The visitors behave differently to ordinary search visitors once they arrive, and that difference is the part worth acting on. I have been doing this work for 13 years, and this is the first channel shift I have seen where the honest numbers are less impressive than the coverage around them, and more useful than the coverage too.

Before anything else, three words that get used as though they mean the same thing: a mention, a citation and a click.

Does AI search send traffic to websites

The short answer, with numbers

It sends traffic, it is measurable, and it is a small share of total sessions. The clearest published example I can point to is a Semrush study reported by PPC Land. Semrush published the study on September 8, 2026, and it found that AI assistants and Google's conversational search surface sent 0.48% of all sessions to manufacturing and industrial websites between January and July 2026, a period in which generative summaries spread across 57% of the tracked searches for that sector.

State the sector and the period whenever you quote that figure, because it is a single vertical measured over seven months, not a universal number. I have seen it recycled as though it described every industry, and it does not.

Mentions, citations and clicks are three different things

A mention is your name appearing in an answer with no link. A citation is the answer linking to a source. A click is a person actually leaving the answer surface and landing on a site. They are counted differently, they mean different things commercially, and most vendor reporting blurs them together. If a tool reports that your brand appeared in 400 answers this month, that tells you nothing about traffic, because a mention without a click produces no session in your analytics at all.

How much traffic AI assistants actually send

What measured studies show so far

The manufacturing study gives the full channel breakdown, which is the most useful part of it. The clickstream measurement of the ten industrial categories produced this share of sessions between January and July 2026: direct at 56.65%, organic search at 22.28%, referral at 12.89%, email at 3.52%, paid search at 2.58%, organic social at 1.33%, AI assistants at 0.45%, paid social at 0.19%, display advertising at 0.06%, and AI Mode at 0.03%.

A second dataset, reported in the same article, points the same way: "Datos placed AI tools below 2% of desktop visits in the first quarter of 2026 while Google held 94% of search." Treat that one as corroboration of a third party panel rather than as a primary figure, but it does not contradict the first study, and two independent measurements agreeing is the most you can ask for at this stage.

Why the share differs so much by industry

The categories where AI referrals show up most are the ones with heavy research behaviour before a decision: software, health, finance, and anything where people compare options over weeks. Manufacturing and industrial buyers research in that way too, which is why a 0.48% figure for a single vertical is not the whole story. Categories where people buy from a local business on a phone see far less of this, and the honest advice for those businesses is to watch their own referral data rather than a headline number.

Small share, different kind of visitor

Conversion value compared with organic search

Small does not mean unimportant, and there is data supporting that. Semrush's study of the impact of AI search on SEO traffic states: "We have seen that the average AI search visitor (tracked to a non-Google search source like ChatGPT) is 4.4 times as valuable as the average visit from traditional organic search, based on conversion rate."

That is Semrush's own dataset for digital marketing and SEO related topics, so treat it as directional rather than universal. But it is the strongest available argument for paying attention to a channel this small, and it matches what I have seen on client accounts: fewer sessions, much later in the decision, and closer to a conversation with a sales team than a click on a headline.

Which pages tend to attract AI referrals

Two patterns show up in the referral data. The first is what those answers link to: Semrush reports that "Our research shows that 50% of links included in ChatGPT 4o responses point to business/service websites (even though sites like Quora and Reddit perform best on an individual domain basis)." Commercial sites are a large share of the outbound links, which is good news for anyone selling something.

The second pattern is about which pages get referenced at all. Semrush's data also notes that "When ChatGPT search cites webpages, the pages it cites rank in traditional organic search positions 21+ for related queries almost 90% of the time, according to our data." Read that as a referral pattern rather than as an instruction: the pages being linked to are frequently not the pages ranking first, which means visibility in answers is not simply a function of your best organic position.

What happens to clicks when an AI answer appears

Click behaviour with and without a summary

This is the effect that is larger than the referral percentage suggests. Pew Research measured what happens when a summary appears above the results: "Users who encountered an AI summary clicked on a traditional search result link in 8% of all visits." And for the comparison group: "Those who did not encounter an AI summary clicked on a search result nearly twice as often (15% of visits)."

Why the effect is larger than the referral number suggests

AI referral traffic is a fraction of one channel. The behavioural change is happening across search as a whole, because summaries appear on queries that people used to click through. A business measuring only AI referrals will underestimate what is happening to its organic traffic, and then look for a ranking explanation for a decline that has nothing to do with rankings. That misdiagnosis is expensive, and I see it more often each quarter.

Finding AI traffic in your own analytics

Referral sources worth checking in GA4

Start with referral traffic and look for sources that map to AI surfaces. Some arrive with recognisable referrers and are easy to isolate. Set up a segment or a report filtered on those sources, and save it, because you are going to want to compare the same view every month rather than rebuilding it.

The direct traffic misattribution problem

Some AI referrals arrive without a clean referrer and are counted as direct. That is not something you can fully fix, and chasing a perfect number in month one is a waste of effort. What matters more is consistency: measure the same way every month and watch the direction. A direct traffic share that creeps upward while your organic clicks fall is worth investigating, and AI surfaces are one of several explanations, alongside email clients and messaging apps that strip referrers.

Building a baseline this month

Set the baseline now, before the numbers are big enough to feel important. Record total sessions, sessions from referral sources that map to AI surfaces, and the conversion rate of each group. In three months, that record will tell you whether this channel is growing in your market, and it will do so with your own data rather than a vendor statistic. If you want a structured way to build that view, the AI search visibility guide walks through it step by step.

What to do with AI visitors when they arrive

Answering the question they arrived with

Someone arriving from an answer surface is usually already part of the way to a decision, and they arrived with a specific question. A page that restates the question in its first sentence and answers it immediately will hold them. A page that makes them scroll past an introduction about the company will lose them, and they will go back to the answer they came from.

Turning a research visit into an enquiry

These visitors are reading, not browsing, so the next step has to be obvious and low friction: a clear statement of what you do, where you do it, and one way to make contact. If the page is a service page, the contact route belongs on it rather than only on a generic contact page. This is a page problem far more often than it is a channel problem, and it is the part I would fix first.

Pages that are likely to be the landing point

Your landing points will usually be the pages that answer questions plainly: comparisons, pricing explanations, service definitions, and posts that resolve a specific problem. Open your analytics, find the pages that already collect direct and referral traffic with no campaign attached, and make sure those pages are ready for a reader who has been given a partial answer and wants the rest. The work on visibility in AI answers starts with exactly this list.

What this means for your search priorities

Keep the fundamentals, add clarity

Nothing about AI answers justifies abandoning the basics. Pages that answer a question clearly, facts that are current, information that is specific, and a site that is easy to crawl are the same things that have always worked. The addition is precision, because a system extracting a sentence cannot infer what you left implied. Write the answer where a reader expects to find it, state the numbers and the dates, and name the thing you are describing.

Content that is worth a link in an answer

Referenced pages tend to contain something a summary needs: a definition, a figure, a comparison, a process. If a page contains only persuasion, there is nothing to extract. That is a useful test for any page you are about to publish, and it is the same test a human reader applies when deciding whether a page was worth opening.

Measuring AI visibility over time

Track referrals, mentions and citations separately

Keep three numbers rather than one. Referrals come from your analytics and are the only ones tied to actual sessions. Mentions and citations come from tracking tools and manual checks, and they describe presence rather than traffic. When someone sends you a report showing a large rise in AI visibility, ask which of the three it is measuring, because two of them do not pay the bills.

What to ignore while the data is thin

Ignore month to month movements in a channel this size. A rise from 0.4% to 0.6% of sessions is not a trend, it is noise from a small sample. What is worth watching is the direction across two or three quarters, and any change in the behaviour of your organic traffic that coincides with summaries appearing on your core queries. There are more posts on search and measurement if you want to go further into the reporting side.

Where this is heading

What to watch in the next twelve months

Watch three things: whether AI referral share grows in your category, whether your organic click through rate declines on queries where summaries now appear, and whether the pages being referenced in your market are pages you own. Any of those turning in the wrong direction is worth a conversation. None of them require panic in a single month.

Setting a defensible expectation for your business

Expect this channel to stay small for a while and to matter more than its share suggests. One projection worth knowing about, clearly labelled as a projection rather than a measured outcome: Semrush concludes from its own data that "AI search traffic has the potential to overtake traditional organic search traffic within the next two to four years." I would not plan a budget on that sentence. I would make sure the pages most likely to be the landing point are ready, and set the baseline that lets me judge it with my own numbers. If you want to talk about what your analytics already show, that is a sensible place to start, and you can talk through what your analytics show before spending anything on changes.

AI search traffic questions I get asked most

So does AI search actually send visitors to websites?

Yes. The traffic is real and measurable, and it is currently a small share of total sessions. The most useful published example is a Semrush study reported by PPC Land, which found AI assistants and Google's conversational search together sent 0.48% of sessions to manufacturing and industrial sites over seven months of 2026. That is a channel you should be able to see in analytics, not one that has replaced organic search.

How do I see AI traffic in Google Analytics?

Look at referral traffic and check for sources that map to AI surfaces, and accept that some of it will be invisible because it arrives without a clean referrer and gets counted as direct. Building a monthly baseline of what you can see is more useful than trying to make the number perfect in month one. The point of the baseline is to know whether the number is growing, not to have a precise figure today.

Is AI traffic more valuable than normal traffic?

There is evidence pointing that way. Semrush reported that the average AI search visitor is 4.4 times as valuable as the average traditional organic visitor based on conversion rate, for its own topic set. Treat that as directional evidence from one dataset rather than a universal multiple, and test it against your own conversion data once you have enough traffic to compare.

Should I change my strategy because of AI search?

Not the fundamentals. The pages that get referenced in AI answers tend to be the same pages that answer a question clearly and are trustworthy, which is the same standard good pages have always been held to. What changes is the incentive to write plainly, answer the question early, and keep facts current, because a machine extracting a sentence cannot infer what you left implied.

Will AI answers take my traffic away?

For some query types, probably. Pew Research found that users who encountered an AI summary clicked on a traditional search result in 8% of visits, while those who did not clicked nearly twice as often at 15%. That is a genuine reduction in clicking behaviour. The response is not to abandon search, it is to make sure the traffic that does arrive lands on a page that converts, and to be present in the answer surfaces as well as the results.

How long before this matters for a small business?

It depends on your category. Categories with heavy research behaviour such as health, finance, trades comparisons and software are seeing more AI mediated discovery than categories where people buy from a local business on a phone. Watch your own referral data each month and let your own numbers, not the industry headlines, decide how much attention it deserves.