AI Referral Traffic Statistics 2026: What the Data Shows and Where It’s Headed

Updated October 22, 2026 · Based on 15 sources

AI assistants sent 0.43% of all website visits in September 2026, or about 4 in 1,000 by our math, across more than a hundred thousand sites in Ahrefs‘ Web Analytics benchmark.

Five of the six AI referral traffic statistics in our table land under half a percent. They differ mostly by which sites were measured and when, and every one misses some AI visits that analytics never labels as AI.

AI referral traffic statistics: how much traffic comes from AI in 2026?

From under 0.2% to 1.08% of visits, depending on whose sites were measured and when. The lowest figures are mostly the oldest. The two largest current panels, Ahrefs (September 2026) and SE Ranking (early 2026), both land under half a percent.

Study	AI share of visits	Method	Sample	Period	Source
Ahrefs Web Analytics benchmark	0.43%	Pooled share of all visits	115,254 sites, worldwide	September 2026	Ahrefs
SE Ranking	0.32%	Pooled share of all visits	101,574 sites, worldwide	January to April 2026	SE Ranking
Ahrefs study	0.25%	Average of each site’s share	81,947 sites, worldwide	June 2025	Ahrefs
Contentsquare benchmark	0.2%	Share of all sessions	6,000+ sites	2025	Contentsquare
Kaiser & Schulze (peer-reviewed)	Under 0.2%	Share of total traffic	973 online stores	August 2024 to July 2025	Frankfurt School of Finance & Management (Kaiser & Schulze, Marketing Science)
Conductor	1.08%	Share of all visits, monthly average	1,215 U.S. enterprise domains	May to September 2025	Conductor

Time explains much of the spread. The lowest figures, from Kaiser & Schulze’s peer-reviewed study of online stores and from Contentsquare, cover 2024 and 2025, when AI’s share was still climbing fast on SE Ranking’s panel. The highest comes from Conductor’s sample of large U.S. companies that use its tools.

Neither Ahrefs’ panel nor SE Ranking’s is a random sample of the web. Both are sites using the vendor’s own tools, and neither is built from local businesses.

In the same Ahrefs data, Google alone sent 23.96% of visits. If you’re weighing AI against search for your budget, our post on whether SEO is still worth it looks at what AI answers are doing to search clicks.

Divide your AI sessions in GA4 by total sessions. As a rough yardstick from Ahrefs’ pooled figure, a site with 10,000 sessions a month would see about 43 identified AI visits by our math. Your own share depends on your industry and audience, so a lower number is a reason to check your tracking first.

Why do AI traffic studies disagree?

They measure different sites, months, countries and ways of counting. Conductor’s figure comes from U.S. enterprise customer domains in mid-2025, while SE Ranking’s covers more than a hundred thousand sites worldwide in early 2026. Averaging them would mix very different businesses, so we keep each number with its sample.

Sample: which sites are in the panel

Industry moves the number a lot, even inside one enterprise sample. In Conductor’s 2025 enterprise data, IT sites got 2.80% of traffic from AI referrals and Consumer Staples 1.91%, while Communication Services got 0.25% and Utilities 0.35%.

Every site in that sample is a large company. That fits an industry effect, though Conductor’s client mix could play a part too.

Geography: where the visitors live

SE Ranking measured AI’s 2026 share at 0.29% of website traffic in the US, 0.27% in the EU and 0.16% in the UK. The worldwide figure sits above all three, so other regions must run higher. Country mix explains only part of the gap between studies.

Method: pooled share or per-site average

A pooled share adds up every visit across all sites, so a few huge sites can set the number. Ahrefs’ 2025 study averaged each site’s own share instead, so a small site counts as much as a big one. Similarweb counts visits and skips share.

Compare your number to the study whose sites look most like yours, and note its method and month. If none fits, your own trend over a quarter tells you more.

What counts as AI referral traffic, and what does Google Analytics miss?

AI referral traffic is a visit that starts with a click on a link in an AI assistant’s answer, such as ChatGPT or Gemini. Clicks from Google’s AI Overviews and AI Mode count as Google search. GA4’s AI Assistant channel also misses many ChatGPT clicks, so analytics counts for AI are a floor.

Google Analytics Help says GA4’s AI Assistant channel covers assistants such as ChatGPT, Gemini, DeepSeek, Copilot and Grok. The rule is narrow. GA4 sets a session’s medium to ai-assistant only when the referrer matches its list, and the channel takes only that medium.

Google’s AI Overviews and AI Mode count as Google search

GA4’s AI Assistant channel leaves out Google’s own AI features. A click from an AI Overview or AI Mode arrives as Google organic search, and Conductor says it left AI Mode out of its AI numbers for that reason.

Google Search Central says Google counts clicks from AI Overviews and AI Mode inside overall Google search traffic in Search Console, a separate system. Our guide to AI Mode in Search Console covers what you can and can’t see.

Where ChatGPT visits go when GA4 can’t label them

Adobe’s Cedric Huesler, as reported by Search Engine Journal, says most ChatGPT clicks arrive with no referrer at all. The utm_source=chatgpt.com tag on ChatGPT’s links is what identifies them.

That tag carries a source but no medium. GA4 records the visit with chatgpt.com as its source, but with no matching referrer it falls outside the AI Assistant channel, usually under Unassigned.

Per MDN Web Docs (Mozilla), a link marked noreferrer tells the browser to drop the Referer header. A click with no tag and no referrer gives GA4 nothing to go on, so it lands in Direct. No study we found measures how many AI visits end up there.

• AI Assistant: sessions whose referrer GA4 matched to an AI assistant.
• Unassigned: tagged ChatGPT clicks with no referrer. The source still reads chatgpt.com.
• Referral: where AI visits sat before the channel existed, and where some custom channel groups still put them.
• Direct: clicks with no tag and no referrer. A sudden jump here can point to an event like the October bug below.

In GA4’s Traffic acquisition report, note your AI Assistant sessions. Then set the primary dimension to Session source, add Session default channel group as a second dimension, and filter for chatgpt.com and the other assistants’ domains. Add only the rows that sit outside the AI Assistant channel. That sum is your AI count, and it’s still a floor.

How fast is AI referral traffic growing?

Fast from a small base, though growth in AI’s share has slowed. SE Ranking measured it climbing from 0.02% in 2024 to 0.24% in 2025, then more slowly into 2026. Visit volume still more than doubled, with Similarweb counting 117.4% more AI referral visits in the year to May 2026.

What that growth looks like in plain numbers

• By our math, SE Ranking’s shares equal 2 in 10,000 visits in 2024 and 32 in 10,000 in early 2026.
• In absolute terms, that’s a rise of 0.30 percentage points.
• The 2024 figure is rounded and comes from before SE Ranking’s regular tracking began, so treat it as rough.

Most of that growth came early. On the same math, SE Ranking’s share grew 12x from 2024 to 2025. SE Ranking reports it grew about 36% from 2025 to 2026; the rounded shares give about a third.

Its monthly series peaked at 0.3511% of traffic in October 2025, and the January-to-April 2026 average is still below that peak. AI’s share on that panel has been roughly flat since.

Visit counts tell a different story. Similarweb found an average of 770.7 million AI referral visits a month worldwide from June 2025 to May 2026, and all seven industries it tracked grew by at least 115% year over year. Share and volume can move apart because they measure different things on different panels.

Similarweb’s fastest-growing category, Beauty (+312.5%), added only about 1.5 million monthly AI visits, from 473.3K to 2.0 million, while News and Travel’s 115.6% growth added about 23.9 million each.

Track your AI sessions each month and look at the visit count next to any percentage change. Doubling a handful of visits still leaves a handful.

Which AI assistants send the most website traffic?

ChatGPT, by a wide margin. In SE Ranking’s 2026 data, ChatGPT sent 74.78% of AI referral traffic, Gemini 11.56%, Perplexity 7.23%, Copilot 3.51% and Claude 2.62%. Conductor’s 2025 enterprise sample put ChatGPT at 87.4%, a gap that may reflect timing as much as sample.

SE Ranking measured Gemini’s referrals up 231% and Claude’s up 320% from 2025 to 2026, while ChatGPT’s own referrals grew 27%.

Statcounter Global Stats, which counts referrals to sites running its tracker, put ChatGPT at 80.48% of AI chatbot referrals worldwide in September 2026.

Leads follow a similar order in the home-services network tracked by SearchLight. In September 2026, 88% of SearchLight’s AI-referred home-services leads came from ChatGPT and 7% from Gemini, with Perplexity and Copilot about 2% each and Claude under 1%.

An assistant has to name your business before anyone clicks, and that’s the step our guide on how to rank in AI answers covers.

Report ChatGPT and Gemini as separate sources each month. ChatGPT sends most AI visits in every source above, so a change in how it tags links will move your AI line the most.

Why did ChatGPT traffic drop in October 2026?

Most likely a tracking bug. Adobe’s Cedric Huesler, tracking about 500 pages with heavy ChatGPT traffic, reported that from about October 8 many ChatGPT links lost their utm_source tag, so those visits showed up as Direct. He says OpenAI confirmed the bug to him privately and a fix began October 15; OpenAI has not commented publicly.

Adobe’s Cedric Huesler, comparing Oct 9-13 with the week before on about 500 pages where ChatGPT drove 5%+ of visits, saw ChatGPT visits fall about 31,000 a day while Direct rose about 26,000 a day. Those pages were picked for heavy ChatGPT traffic, so the daily totals won’t scale to a typical site.

Per Huesler, the fix began early on October 15, and by 9 a.m. PT tagged ChatGPT clicks were back to roughly 95% to 100% of the same hour a week earlier.

Leads attributed to ChatGPT fell too. In SearchLight’s home-services network, they dropped 26% between the weeks of September 28 and October 5, then partly came back.

How to check your October report

1. In GA4’s Traffic acquisition report, compare October 8 to 14, 2026 with October 1 to 7.
2. For each week, add your chatgpt.com sessions to your Direct sessions. If chatgpt.com fell while that total held, the bug likely moved those visits into Direct.
3. Check that chatgpt.com sessions came back after October 15.
4. Expect leads attributed to ChatGPT to look low that week too. Don’t treat it as a real drop in demand.
5. On a small site with a handful of AI visits a week, the shift won’t show against normal swings in Direct. Just annotate October 8 to 14 in GA4 and move on.

If chatgpt.com sessions fell and Direct rose by a similar amount that week, the bug is the likely cause. Annotate the week and keep your plans as they were.

Where could AI referral traffic be by 2028? A labeled scenario

We found no published forecast, so we ran two separate calculations. If AI’s share kept growing at the yearly rate SE Ranking measured, its early-2026 reading would reach about 0.6% around early 2028. Separately, at Similarweb’s growth rate, AI referral visits would reach 4.7x their October 2026 level by October 2028.

Each path asks what happens if one measured rate holds. They use different measures, so we keep them apart and don’t rank them.

Share path: SE Ranking’s data only

SE Ranking’s January-to-April 2026 share, in the first table, would reach the figure above about two years later if it kept growing about 36% a year, the rate SE Ranking reports from 2025 to 2026.

Treat it as a what-if. SE Ranking’s monthly series has been roughly flat since its October 2025 peak, so this shows what the year-over-year rate would mean if it held. It isn’t a trend visible in today’s data.

Volume path: Similarweb’s growth rate

Similarweb measured AI referral visits growing 117.4% a year, and the table applies that rate for two years from an illustrative October 2026 index. It counts visits only. If other traffic grows too, AI’s share would rise far less, so we don’t turn it into a share.

Date	AI referral visits, indexed (Similarweb rate)	Source
October 2026 (illustrative start)	100	Enleaf calculation
October 2027	217	Enleaf calculation
October 2028	473	Enleaf calculation

Why these paths could be wrong

• Growth is uneven by region. SE Ranking found the EU’s AI share fell a relative 12% from February to April 2026.
• Similarweb’s data ends in May 2026, so the October 2026 starting point is our assumption.
• Single months can swing. Similarweb’s monthly data shows AI referrals to marketplaces jumping 36% month over month in May 2026.
• The share path could also undershoot. If search traffic keeps falling, as it did by about 21% in a year for the average site in a 2025 Ahrefs study, AI’s share would rise faster than its visits.

Plan for AI as a growing minority channel over the next two years, and keep investing where your leads come from today. Check your AI count each month.

What should your business measure?

Four things: AI-referred sessions, AI’s share of all sessions, the leads and sales they produce, and the mix by platform and landing page. Track volume and share together, since share can rise while AI visits stay flat if your other traffic falls.

Measure	Where to find it in GA4	What it tells you
AI-referred sessions	AI Assistant row, plus chatgpt.com and other assistant sources in other channels	How many visits AI assistants send you
AI share of sessions	AI-referred sessions divided by total sessions	Whether AI is growing faster than your other traffic
Leads and sales by channel	Key events, such as form fills and tracked calls, by channel	Whether AI visitors turn into customers
Platform and landing page	Session source, then landing page	Which assistants send people, and where they land

If the AI Assistant row is missing, fix your Google Analytics setup first. Our AI search optimization services start from this same baseline.

What local businesses are seeing

Home-services firms are the only small-business benchmark we found. In SearchLight’s network of HVAC, plumbing, electrical and roofing firms, AI assistants drove 0.24% of web leads in September 2026, and that lead share rose 85% from January.

In the same network, 63% of AI-referred home-services leads came in as form fills and 33% as calls, versus 79% calls and 16% forms across all web leads (September 2026). A call placed straight from an AI answer, with no click, can’t be credited to AI, so AI-attributed leads lean toward forms by design.

BrightLocal‘s 2026 survey of 1,002 U.S. adult consumers found 45% had used ChatGPT or another AI tool for local business recommendations. Analytics only sees the click, and many of those choices happen with none, the pattern behind AI and zero-click searches.

To see whether assistants name you at all, track your AI visibility separately: ask them the questions your customers ask, on a schedule, and log whether you show up.

Do AI visitors convert better?

Not reliably. In Kaiser & Schulze’s peer-reviewed study of online stores (August 2024 to July 2025), ChatGPT visits converted better than paid social but lower than all other established channels.

Adobe Digital Insights measured something different: all AI visits against all other traffic to U.S. retail sites. There, AI visits converted 42% better in March 2026. Neither study covers your site.

Measure leads alongside visits. Set up form fills and tracked calls as key events, then compare AI-referred leads with your other channels once a quarter.

Frequently Asked Questions

What we’d do with these numbers

For most small and local businesses, AI referral traffic isn’t yet big enough to move budget away from search. Measure it well now: count chatgpt.com sessions outside the AI Assistant channel too, and give each AI source its own line in your reports.

Revisit that call once AI-referred leads hold a steady few percent of your total for two quarters, or when your AI share keeps climbing while search falls. That’s the point to build content aimed at getting named in AI answers.