ChatGPT Owns 92% of AI Referral Traffic: What the 2026 Data Means
On July 6, 2026, the SEO firm Previsible published its third AI Traffic Study, and the headline number is hard to ignore: ChatGPT now drives 92.4% of all trackable large-language-model (LLM) referral traffic. The report analysed 6.77 million AI-driven sessions across 166 real websites between November 2024 and May 2026 — one of the largest public datasets on where AI-referred visitors actually come from.
If you have been treating "AI search" as a diffuse cloud of ChatGPT, Gemini, Perplexity and Copilot, the data says otherwise. Standalone AI referral traffic in 2026 is, for now, almost entirely a ChatGPT story — and that concentration changes how you should prioritise your AI search optimization work. Here is what the numbers show and what to do about them.
What the Previsible study actually measured
The study is worth understanding before you act on it, because its scope defines what the 92.4% figure does and does not cover.
- Sample: 166 GA4 properties across nine verticals — SaaS, e-commerce, finance, legal, health, insurance, education, publishing and ticketing.
- Volume: 6.77 million LLM-referred sessions over 19 months (November 2024 – May 2026).
- What counts: only referral visits where a user clicked a link inside an AI product and landed on a site. Total monthly LLM sessions in the sample grew from 65,249 to 644,478 — a 9.9x increase.
- What is excluded: Google AI Overviews. Because those clicks are attributed to normal Google organic traffic, they are invisible here — and Previsible notes they likely dwarf standalone LLM referrals combined.
So the 92.4% share describes the standalone chatbot slice of AI traffic, not all AI-influenced search. That caveat matters, and we return to it below.
The market share breakdown
Within that standalone slice, the concentration is extreme. ChatGPT is not merely leading — it is nearly the entire market, and the gap is widening, not closing.
| Platform | Growth over study | Where it stands in May 2026 |
|---|---|---|
| ChatGPT | 12.8x | 92.4% share; the dominant referrer by a wide margin |
| Gemini | 3.2x | Quiet but steady #2 (≈18,000 monthly sessions) |
| Claude | 64x | Fastest grower; overtook Perplexity in March 2026 |
| Perplexity | Down 61% | Peaked March 2025, has declined since |
| Copilot | Down 96% | Collapsed from its 2025 peak to a rounding error |
Two sub-plots stand out. First, Claude grew 64x — from 133 sessions in November 2024 to 8,528 in May 2026 — and now sends more referral traffic than Perplexity, especially to developer, SaaS and professional-services sites. Second, Copilot effectively withdrew from the referral picture, dropping roughly 96% from its peak. The story of 2026 is consolidation: one giant, one steady runner-up, one fast climber, and a long tail that is thinning out.
Why this is fragile, not permanent
A 92% share sounds like a moat, but the same dataset contains a warning. In November 2025, a single ChatGPT product change cut its referral volume roughly in half within one month — from 448,412 sessions to 213,345. Traffic recovered, but the episode shows how exposed you are when one company controls almost all of a channel. A tweak to how ChatGPT renders links, cites sources, or keeps users inside the chat can move your numbers overnight, with no algorithm announcement and no appeal.
The lesson is not "ignore ChatGPT" — it is "do not build a business that depends on a single referrer you do not control." Concentration is an opportunity and a risk at the same time.
Where AI traffic actually lands on your site
One of the more actionable findings has nothing to do with market share. Previsible looked at which pages AI-referred visitors reach, and the pattern surprised many marketers:
- ~25% of AI-referred traffic site-wide lands on internal search-results pages — and for ChatGPT specifically, that figure is 28.8%. Users arrive mid-task, ready to keep searching.
- E-commerce sites see 43% of their AI traffic hit product pages. AI tools are increasingly sending people straight to a purchasable item rather than a top-of-funnel article.
- Penetration varies wildly by vertical: e-commerce grew 37x, insurance 18.9x (the highest penetration of any sector), education 5.4x. Health was the only vertical that declined.
The takeaway: AI referrals behave less like blog readers and more like intent-loaded shoppers. If your internal search and product pages are slow, thin or hard to parse, you are wasting the most valuable visitors AI sends you.
What to do with this data: five priorities
1. Optimise for ChatGPT first
When 92% of standalone AI referrals come from one product, that is where your first hour of effort belongs. Make sure ChatGPT's crawler (OAI-SearchBot and GPTBot) is allowed in your robots.txt, keep key facts in server-rendered HTML rather than JavaScript, and write answer-first passages that are easy to quote. Our full guide to optimizing for AI search covers the citation mechanics in detail.
2. Treat internal search as an acquisition channel
If a quarter of AI visitors land on your search-results pages, those pages are no longer just navigation — they are landing pages. Make them fast, index-friendly where appropriate, and capable of guiding a cold visitor toward a conversion. Empty-state and zero-results pages deserve real design attention.
3. Make product and pricing data machine-readable
AI tools favour sources they can parse cleanly. Put prices, specs and availability in the HTML (and in Product schema) instead of hiding them behind "contact us" forms or images. Machine-readable pricing is now a discoverability feature, not just a UX nicety. Getting this right is largely an engineering job — retrofitting valid Product structured data and fast, server-rendered templates across a large catalogue is the kind of build a web development studio like YuSMP Group handles day to day.
4. Watch Claude as an early signal
Claude's 64x growth is small in absolute terms but fast, and it skews toward technical and professional audiences. If you sell to developers, agencies or B2B buyers, Claude referrals are a leading indicator worth tracking now — before the volume becomes obvious to everyone.
5. Measure by page type, not site-wide averages
A blended "AI traffic went up 10%" number hides the real story. Segment AI referrals by landing page and vertical so you can see whether product pages, guides or search results are pulling the weight. Google's new reporting tools help here — see our walkthrough of the Search Console AI performance report.
Don't forget the biggest player isn't in the chart
The study's most important sentence is a footnote: it excludes Google AI Overviews. Those AI answers appear above billions of Google searches, and their clicks are counted as ordinary organic traffic — so they are absent from the 92.4% figure entirely. In reality, Google remains the centre of AI-influenced discovery; ChatGPT simply dominates the standalone chatbot referrals that are easy to isolate in analytics.
Practically, that means classic SEO fundamentals still carry most of the weight. The same authoritative, well-structured pages that earn Google rankings are the ones AI Overviews and ChatGPT both pull from. Recent ranking volatility — including the June 2026 spam update — is a reminder that the organic foundation underneath all of this is still moving.
The takeaway
The 2026 data tells a clear story: standalone AI referral traffic is real, growing fast, and overwhelmingly funnelled through ChatGPT, with Claude the one to watch. But the smart response is not to abandon everything for a single chatbot. It is to optimise for ChatGPT first, harden your product and internal-search pages for high-intent AI visitors, keep your data machine-readable, and never forget that Google AI Overviews — invisible in this report — likely send even more. Concentration is a gift and a hazard. Take the traffic, but don't bet the company on one referrer you can't control.
