ChatGPT commands 92% of AI referral traffic. Here it shows 6.77 million times.

For the past twelve months, the industry has been betting on which AI platform will win adoption. Confused looked like a native opponent of search. The pilot looked like a corporate Trojan horse. There is no betting.

Visual (Disclosure: I am the CPO and co-founder) recently published its third AI Traffic Study, analyzing 6.77 million LLM-driven sessions. The data show assocd. LLM’s monthly sessions grew 9.9x, reaching 644,478 in May 2026. And 92.4% of that traffic came from one site.

The plateau was a break

By mid-2025, AI traffic seems to be booming in other sectors. It wasn’t like that.

Times rose from 65,249 in November 2024 to 396,278 in August 2025, then fell further in November 2025, before hitting new highs of 428,203 in February 2026 and 644,478 in May.

That November dip needs context.

Sessions fell 50% in one month, driven almost entirely by ChatGPT referrals falling from 448,412 to 213,345. Some platforms are stuck. This was probably a model-related change; we’ve seen modest product changes dramatically swing referral traffic, like last fall when many sites lost a portion of their ChatGPT traffic because the model started favoring Wikipedia and Reddit. Sessions returned to 442,609 in December.

The lesson: one vendor’s product decisions can slow down your AI traffic overnight. Plan a tense situation.

Integration, not competition

At last publication in December 2025, ChatGPT held approximately 84% of the shares, followed by Perplexity at 8.9%, Gemini at 4.5%, Copilot at 2.1%, and Claude at 0.6%. After six months, the field has fallen towards the leader.

In the full dataset, ChatGPT commands 92.4% of trackable LLM referral traffic, growing 12.8x over 19 months with no sign of slowing. It is the only LLM that sends a reasonable volume of referrals at scale. Configuring “AI visibility” without prioritizing ChatGPT means anonymous configuration.

Important note: this measures the traffic of independent LLM referrals. AI discovery within Google results, including AI Overview, almost certainly drives more AI traffic than all independent platforms combined, but it operates on a different measurement paradigm and is not covered here.

The challengers turned

The surprise is not at the top. He is the one who goes under.

Claude

Claude grew 64x, from 133 sessions in November 2024 to 8,528 in May 2026, and surpassed Perplexity in March 2026 for the first time. It sat first.

Claude was down by 2025, then accelerated 4x in two months as his agent tools and business integration were adopted. The business advantage the industry expected Copilot to win may be seen in Claude instead.

If your audience includes buyers of technology, developers, or professional services, Claude’s visibility is now important, and the early booking window is open.

Gemini

Gemini is a quiet number two: 3.2x growth with virtually no volatility. Its Desktop and Android integration means that referral numbers may be lower than its actual adoption.

Confusion & Copilot

Freaking peaked at 17,507 monthly sessions in March 2025 and has fallen 61 percent since then. The pilot is down 96% from its August 2025 peak, from 8,651 sessions to 339.

No more betting on traffic acquisition growth. Both are changing to keep users within their own experience: browsers, agents, and methods where they don’t need to send you traffic at all.

Where LLMs are sending users, and why they should change your path

The most effective finding of the study is not market share. Landing pages.

ChatGPT sends 28.8% of its traffic to internal search results pages. Across all industries, about 25% of AI-directed traffic comes from internal searches.

The model trusts your domain but can’t choose the right page, so it sends users to your search box and lets them navigate. This pattern persists across specific periods and periods, suggesting that it is systematic for improved generational returns rather than a temporary quirk.

Think about what that means. The model has done the hard work of choosing your domain. Your internal search UX now determines whether those high-intent visits convert or bounce.

For many sites, internal search is an overlooked navigation feature, not a discovery point. That has to change.

Direct observation tells several different stories:

• SaaS traffic comes from search pages (34.6%).
• Publisher traffic comes to news pages (54%), yet against 120 million+ organic sessions, publisher penetration is 0.11%; publishers produce LLM content they cite and capture almost one of the resulting traffic.
• Ecommerce traffic lands on product pages, with purchase intent already generated.
• Education traffic goes directly to course pages (52%), surpassing marketing content.
• Health traffic remains in About (42.1%), users checking the source before the content.
• Official traffic is spread across the blog, about, contact, and site pages: a complete test arc.

Platforms have personality, too:

• ChatGPT and Gemini search pattern models: domain trust, page-level uncertainty.
• Confusion and Claude are content selection models that select specific pages and index over long form.

If your strategy depends on content planning to drive qualified traffic, Perplexity and Claude have done their share.

What you have to do now

• Configure ChatGPT first. Expand to another area where the volume allows.
• Take care of Claude. Perplexity passed in March. Initial placement combinations.
• Treat product pages as entry points for AI. Product pages capture 43% of LLM e-commerce traffic. Organized, comparable product data is a must-have now.
• Make the price machine-readable. “Contact us for pricing” gives AI systems nothing to summarize, compare, or recommend.
• Prioritize internal search. It’s a discovery tool, not a navigation feature.
• Track AI traffic by page type, not globally. Your site average hides where AI traffic is concentrated. Your pricing page can generate 3x your site-wide traffic.

The next question is one that no one can answer: the conversion rate with the LLM platform. Which platforms send users who buy, and which send users who bounce?

We built this dataset to answer that. If the last 19 months are any indication, the answers will change faster than most parties are ready for.

About the data

166 GA4 properties, November 2024 to May 2026, from SaaS, ecommerce, finance, legal, healthcare, insurance, education, publishing, and tickets. All 166 properties are present throughout the full 19-month window, so the trajectories reflect behavioral change rather than sample expansion.

Report

You can find the full report at predictable.io.