On July 6, 2026, the firm Previsible published its 2026 AI Traffic Report, one of the few large-scale measurements of the traffic AI actually sends to websites. The verdict is stark: ChatGPT concentrates 92.4% of referral traffic from large language models, far ahead of Gemini, Claude and Perplexity. The study is based on 6.77 million sessions measured across 166 Google Analytics 4 properties, between November 2024 and May 2026, according to the report published by Previsible.
In short: when an AI sends a visitor to a website, more than nine times out of ten that visitor comes from ChatGPT. Gemini, Claude and Perplexity share the scraps. For a business that wants traffic and leads, this redefines a simple priority: being cited by ChatGPT is no longer one option among many, it's the number-one AI channel, and by a wide margin.
What the Previsible report says
Previsible tracked LLM referral traffic over nineteen months. Total volume grew nearly tenfold: from around 65,000 monthly sessions in late 2024 to 644,478 sessions in May 2026, a 9.9x increase. But that growth is highly concentrated. Here is the breakdown by platform, as measured in the study:
Two moves are worth noting. Claude grew 64x over the period and overtook Perplexity in March 2026, carried by technical and professional audiences. Copilot, by contrast, collapsed 96% from its August 2025 peak and now sits below 0.5%. AI referral traffic isn't a stable market: it's a hierarchy hardening around one dominant leader.
Why this number is subtler than it looks
Be careful not to conflate two things. This report measures referral traffic: clicks that actually land on a website. It is not AI usage share, nor the number of times your brand is cited without a click. And that is precisely the terrain Google plays on: its AI Overviews and AI Mode most often answer without sending anyone out. They are massively used, but generate little measurable traffic, because they keep the user inside Google.
ChatGPT, by contrast, links out more. Its dominance of referral traffic therefore reflects both its user base and its tendency to send a click. This lines up with what we observe elsewhere: visitors coming from ChatGPT engage twice as much as the average once on the site. Fewer visits, but qualified ones.
What it changes for your visibility
For eighteen months, the dominant narrative has hammered "optimize for Google AI Overviews." That's useful for awareness, but if your goal is to generate visits and leads, this data shows you're chasing a minority of the actually clickable AI traffic. In volume, the priority is ChatGPT.
That is exactly the ground GEO covers: optimization for generative engines. Being cited by an AI without ranking in Google's top 10 becomes a goal in its own right. And citation logic differs from one model to the next: we broke down how reasoning models pick their sources in an earlier article.
The test to run today: open ChatGPT, ask the question a prospect would ask to find a provider like you, and see who gets cited. If it isn't you, it isn't an ad-budget problem, it's a structured, citable content problem. to see where you stand.
What to do now
Structure your content to be cited by ChatGPT
ChatGPT draws on clear, sourced, direct passages. Answers at the top of each section, explicit headings, dated and attributed data. That's the foundation of the method for showing up in ChatGPT and AI Overviews.
Publish what AI cannot fabricate
A model easily generates generic content. It cannot invent your proprietary data, your quantified customer results, your field expertise. That is precisely the material ChatGPT cites, because it exists nowhere else. Adopt a real structured GEO strategy rather than a pile of articles.
Measure your AI traffic and citations
Google Analytics 4 now isolates traffic coming from AI assistants. Track that channel, cross-reference it with how often you're cited in ChatGPT, and make it a baseline as important as your average Google position. You can't steer what you don't measure.
The limits: what this article doesn't cover
Three honest caveats. First, the report measures referral traffic, not awareness: a citation in a Google AI Overview, even without a click, builds credibility that doesn't show in these figures. Second, the Previsible sample is mostly English-speaking and B2B; the exact proportions may differ in other markets, even if ChatGPT's dominance is consistent with global traffic data. Third, these shares move fast: Claude grew 64x in nineteen months. The landscape is stabilizing around ChatGPT today, but it isn't frozen.
Our take
At Cicéro, we've said it since the first AI Overviews: SEO isn't dying, it's splitting in two. This report adds a useful precision to the debate. Optimizing for Google stays essential for awareness and zero-click, but when it comes to actually clickable traffic, only one name really matters today. Those who structure their content to be cited by ChatGPT now, while most are still debating AI Overviews, will build a lead that's hard to catch.
Frequently asked questions
Which AI sends the most traffic to websites in 2026?
Why doesn't Google AI Overviews show up in these figures?
Should I stop optimizing for Google AI Overviews?
How do I know if my business is cited by ChatGPT?
Sources
- → Previsible — 2026 AI Traffic Report — study published July 6, 2026, 6.77M sessions across 166 GA4 properties (Nov 2024 – May 2026).
- → Google — Search updates, I/O 2026 — official announcement of one billion monthly users for AI Mode.
Growth and SEO content strategist, I founded Cicéro to help businesses build lasting organic visibility — on Google and in AI-generated answers alike. Every piece of content we produce is designed to convert, not just to exist.
LinkedIn