How do I know if AI search is driving traffic to my website?
AI referral traffic is small, growing, and systematically under-counted. Finding it takes ten minutes of analytics configuration and changes how you value the whole channel.
AI search sends less traffic than the discussion implies and more than most analytics setups report. Both halves of that sentence matter, and the gap between them is where teams either dismiss the channel or overestimate it.
Finding it in analytics
Direct AI referrals are visible if you look for them. Filter your analytics by referrer domain and check for:
chatgpt.comandchat.openai.comperplexity.aigemini.google.comclaude.aicopilot.microsoft.com
Build a segment or a custom channel group covering all of them so it is reported as one channel rather than scattered across referral rows nobody reads. Ten minutes of configuration, and it turns an invisible channel into a tracked one.
Then look at the landing pages. Which of your URLs are receiving these visits tells you which pages are actually being cited — a useful cross-check against your prompt-level tracking, and occasionally a surprise.
Where it hides
Four reasons the reported number understates reality.
Missing referrers. Some assistants and some client configurations pass no referrer header. Those visits land in direct traffic, where they are indistinguishable from someone typing your URL. This is the single largest source of under-counting.
AI Overviews are not referrals. A click from a citation inside a Google AI Overview arrives as ordinary Google organic traffic. It is not separated out, and for most sites this is a larger volume than all the chat assistants combined — invisible by design.
The delayed visit. Someone asks an assistant, gets your name, and searches for you an hour later or types your URL the next day. That visit is branded organic or direct, and the assistant that caused it appears nowhere.
The visit that never happens. The assistant answered completely and nobody clicked. This is not under-counting — it is genuinely zero traffic — but it is where a great deal of the value sits, because the shortlist was still formed.
The most important thing AI search does for you produces no traffic at all: it decides whether you are on the shortlist.
How AI traffic behaves
Per visit, AI referral traffic tends to be better than organic, and the reason is structural.
Someone arriving from an assistant has usually been given a shortlist of two or three options and chosen to investigate yours specifically. They arrive with the category understood, the alternatives known, and a reason to be looking at you. That is a materially later funnel stage than someone clicking a search result.
The observable pattern on most sites: lower volume, higher pages per session, longer time on site, and a conversion rate above the organic average. Check yours rather than assuming — set up the segment and compare it against organic on the same metrics for the same period.
Connect the traffic to the answers
Klepha tracks which prompts cite you and which pages they point at, so referral spikes have an explanation.
Run my free scanThe indirect signals
Because so much of the effect is invisible to referral tracking, three indirect measures matter.
Branded search volume. When assistants begin naming you, more people search your brand name. Track non-branded and branded queries separately in Search Console — a rise in branded impressions without a corresponding campaign is one of the clearest available signals that AI visibility has improved.
Direct traffic. Rising direct traffic with no offline activity, no campaign and no branded search explanation is worth investigating against your AI visibility trend. The two frequently move together, for the missing-referrer reason above.
Self-reported attribution. Adding “how did you hear about us?” to a form, with an AI assistant option, is crude and currently one of the more reliable instruments available. Sales teams can usually also report whether prospects arrive already shortlisted.
What volume to expect
For most sites in 2026, direct AI referrals are a low single-digit percentage of organic traffic. That figure is growing and varies enormously by category — technical, B2B and research-heavy categories see substantially more than local or transactional ones.
Put that against the scale of the surfaces: Google reported 2.5 billion monthly active users for AI Overviews and 1 billion for AI Mode in May 2026. The audience is enormous and the click-through is small, which is precisely the point. Pew Research found users clicked a result in 8% of visits where an AI summary appeared, against 15% where one did not.
The strategic conclusion is not “AI search does not send traffic, ignore it”. It is that AI search increasingly decides who gets considered, and traffic is a lagging, partial measure of that.
A ten-minute setup
Everything in this article depends on being able to see the channel, which most analytics configurations do not by default. The setup is short.
- Create a channel group or segment covering the assistant referrer domains listed above. Name it something unambiguous like “AI assistants” so it appears in standard reports rather than needing a custom query.
- Add a landing-page breakdown to that segment. Which pages receive AI referrals is the most useful single view, because it identifies what is actually being cited.
- Add conversion rate to the same view, compared against organic. This is usually the number that changes internal opinion about the channel.
- Set a Search Console branded/non-branded split using a regex filter on your brand terms, so the indirect signal is visible alongside.
- Annotate the date you made these changes, because the channel will appear to spring into existence and someone will ask why.
Do it before you start optimisation work rather than after. A channel that only becomes visible halfway through a programme cannot demonstrate that the programme caused anything.
How to value it
Three components, and reporting only the first understates the channel badly.
Direct referrals. Measurable, small, high quality. Value them at their actual conversion rate rather than at your site average.
Attributed branded search. The portion of branded search growth that tracks your AI visibility improvements. Imprecise, real, and worth estimating explicitly rather than ignoring.
Shortlist entry. Unmeasurable directly and the largest component. In categories where buyers shortlist before contacting anyone, being named is qualification. The honest way to represent this is as a competitive metric — share of voice against named competitors — rather than as a traffic number, because that is what it actually is.
Three attribution traps
Attributing all direct-traffic growth to AI. Tempting, unprovable, and it will be challenged. Direct traffic rises for many reasons including offline activity, email, and app referrals. Present the correlation, not a causal claim.
Comparing AI referral volume to organic volume and concluding it does not matter. This compares a channel that produces visits with a channel that mostly produces shortlist decisions. The right comparison for AI search is share of voice against competitors, with referral traffic reported as a secondary benefit.
Assuming zero clicks means zero value. Pew Research found clicks fell from 15% to 8% of visits when an AI summary was present — but the 92% who did not click still read something, and what they read either named you or named a competitor. A channel can influence a purchase without generating a session.
What normal looks like
Because there is no established benchmark, teams frequently misjudge whether their numbers are good. Some rough orientation, with the caveat that variance by category is enormous.
Direct AI referrals at a low single-digit percentage of organic sessions is typical for a mid-sized site in 2026. Conversion rate on that traffic running above your organic average is common and worth confirming rather than assuming. A branded search increase that tracks improvements in AI mention rate is a good sign the channel is working upstream of the visit.
What would be unusual: AI referrals exceeding ten percent of organic, which usually indicates either a very technical audience or a measurement error such as a mis-scoped channel group. Worth checking the definition before celebrating it.
Reporting it honestly
Three rules keep this credible with people who will reasonably ask hard questions.
Report the channel separately. Not folded into referral or direct. A named channel with its own volume, conversion rate and trend.
State the under-counting explicitly. Say that AI Overview clicks arrive as Google organic and that missing referrers land in direct. A number presented with its limitations survives scrutiny; the same number presented as complete does not.
Do not claim attribution you cannot support. The temptation to attribute all direct traffic growth to AI is strong and will be caught. Present the referral number as measured, the branded lift as correlated, and shortlist presence as a separate competitive metric with its own evidence.
One further habit worth adopting: record the date you first saw a citation for each page, alongside the referral numbers. Citations and referrals move together with a lag, and having both dated makes the relationship visible rather than assumed.
Our guide to monitoring SEO and AI performance together covers building one report that carries all three without double-counting.
Sources
- Pew Research Center — approximately 68,000 real search queries; clicks fell from 15% to 8% of visits when an AI summary was present.
- Google — 2.5 billion monthly active users for AI Overviews and 1 billion for AI Mode, stated 19 May 2026.
- Ahrefs — CTR impact of AI Overviews; approximately 58% decline for top-ranking pages (February 2026).
Frequently asked questions
How do I know if AI search is driving traffic to my website?
Filter your analytics by referrer domain for chatgpt.com, perplexity.ai, gemini.google.com, claude.ai and copilot.microsoft.com. Those are direct, measurable AI referrals. Then look for indirect signals: unexplained growth in direct traffic, rising branded search in Search Console, and prospects reporting they found you through an assistant.
Does AI referral traffic show up in Google Analytics?
Yes, as referral traffic from the assistant’s domain, though it is frequently mis-bucketed. Some assistants pass no referrer at all, in which case the visit lands in direct traffic and is indistinguishable from someone typing your URL. That is the main reason AI referral volume is systematically under-counted.
Is AI referral traffic worth much?
Per visit, usually more than organic. Someone arriving from an assistant has typically been given a shortlist and chosen to investigate you specifically, so they arrive further down the funnel. Volume is small relative to organic for most sites; conversion rate is often materially higher.
Why is my direct traffic increasing?
One increasingly common cause is AI referrals that pass no referrer header, which land in direct. If direct traffic is rising without a corresponding campaign, offline activity or branded search increase, check whether your AI visibility has improved over the same period — the two frequently move together.
Can I track which AI answer sent someone to my site?
Not directly. Assistants do not pass query-level information the way search engines historically did, so you can identify the platform but not the prompt. The workaround is to correlate: track which prompts you are cited for, and match increases in referrals against the pages those citations point at.
How much traffic should AI search be sending me?
For most sites in 2026 it is a low single-digit percentage of organic, though it varies enormously by category and is growing. Treat the absolute number as less important than the trend and the conversion rate — the strategic significance of AI search is shortlist entry, which happens before any visit is recorded.