AI traffic sources compared: ChatGPT, Perplexity, Gemini, Copilot, and Claude in 2026

How big is AI-referred traffic in 2026?

AI engines are now a measurable traffic source with a behavior profile of their own. This guide compares the five that matter for e-commerce: how much traffic each one actually sends, which index each one answers from, which crawlers you need to allow, and what the visitors they refer do once they reach your store. Every number here links to the source that published it.

Two things are true at once. As a share of total web traffic, AI referrals are still small. As a growth curve, nothing else in the acquisition mix is moving this fast.

That is the honest counterpoint. The same SparkToro zero-click study found that 68.01% of US Google searches ended without any click at all in early 2026. The picture that emerges is a small but compounding channel arriving just as traditional search sends fewer clicks than ever. What makes it worth optimizing for now is not volume, it is the conversion profile covered below, and the cost of being invisible when the volume arrives. For how to find this traffic in your own analytics, see our GA4 walkthrough.

Which AI engines send the most traffic?

By referrals, ChatGPT is the channel. Across SE Ranking's 101,574-site dataset, ChatGPT accounted for 75.96% of AI referral traffic in 2026 through May, with Gemini, Claude, Copilot, DeepSeek, Grok, Poe, and Mistral splitting the rest. That share was 79.74% in 2025: dominance, but eroding.

Usage is shifting faster than referrals. Per Similarweb's tracking, ChatGPT's share of worldwide gen-AI web traffic fell from about 76% in June 2025 to about 53% by May 2026, while Gemini rose from under 9% to roughly 28% and Claude posted the largest proportional gain of any tracked platform, from barely 2% to close to 9%. Usage share is not referral share, but referrals follow usage with a lag: the engines gaining users are the ones whose citations will matter next.

For scale context, Similarweb notes that ChatGPT still receives more web visits worldwide than Gemini, Claude, and Perplexity combined, and has the highest daily-return stickiness of the four. Perplexity is the smallest of the group by visits, which is worth knowing given how often it is described as a major traffic source.

Where does each engine get its answers?

This is the comparison that actually decides whether you get cited. Each engine retrieves from a different index, which means visibility is engine by engine: being readable to one buys you nothing with the others.

Google's setup deserves a note, because it changed in 2026. There is still no dedicated crawler for AI answers: Google states that a page only needs to be indexed and snippet-eligible in normal Search to appear in AI features, and Google-Extended is a robots.txt control token rather than a crawler. What is new is Google-Agent, a user agent added in March 2026 for AI agents that browse and act on a user's behalf. Like all of Google's user-triggered fetchers it generally ignores robots.txt, so it changes what shows up in your logs, not what you should allow.

OpenAI's documentation is explicit that sites blocking OAI-SearchBot will not be shown in ChatGPT search answers, and Perplexity recommends allowing PerplexityBot for the same reason. The practical work is twofold: allow each engine's crawlers in robots.txt, and serve pages these crawlers can actually read. Most of them do not execute JavaScript, so a storefront that renders as an empty SPA shell is invisible regardless of what robots.txt says. Our AI storefront setup guide covers both steps.

How do AI-referred visitors behave once they land?

The reason this channel punches above its volume: the research already happened inside the conversation, so the visitor who clicks through arrives with intent formed.

Two caveats we would want a reader to know. Adobe's conversion advantage is a monthly reading that moves: it was 42% in March 2026 and 54% in May, and in March 2025 AI traffic still converted 38% worse than average, so the advantage itself is new. And Semrush's 4.4x was measured on digital marketing topics, not retail, so treat it as directional for e-commerce.

One more Semrush finding reframes the whole game: ChatGPT cites webpages ranking in positions 21 and beyond in traditional search nearly 90% of the time. Ranking on page one of Google and being cited by an AI engine are different competitions with different winners; GEO vs SEO maps where the two diverge. For what that means in practice, see how to rank in ChatGPT for e-commerce.

Why static pages lose AI-referred shoppers

Here is the mismatch. A shopper who arrives from ChatGPT or Perplexity was mid-conversation: they asked a specific question, got a specific recommendation, and clicked through to verify it. Adobe's data says they will give you more time and more pageviews than any other visitor. A static landing page built for a Google visitor answers none of the question they arrived with, and the conversation that understood them stays behind in the chat window.

That is the gap between being cited and being bought from. Winning the citation gets the click; what the visitor finds on arrival decides whether the 54% conversion advantage lands on your store or evaporates. This is the problem agentic storefronts exist to solve: meeting an AI-referred visitor with the answer they came for instead of a hero banner.