chatgpt's 92% ai traffic number is real. it's also not the number that matters if you're a solo operator.
a study came out this month with a number that sounds like a verdict: ChatGPT drives 92.4% of AI referral traffic. If you've been unsure where to put your AI visibility effort, that reads like a clear answer, go all in on ChatGPT. It's the wrong takeaway, and the study itself says so if you read past the headline.
AI referral traffic is defined as website visits that arrive directly from an AI assistant after it named or linked your site in an answer. It excludes AI-influenced visits that pass through a normal search click, which is why standalone-assistant measurements understate the total AI effect.
what the study actually measured
Previsible, an SEO agency that's tracked AI referral traffic for two years, analyzed 6.77 million AI-driven sessions across 166 sites from November 2024 through May 2026. Across that dataset, ChatGPT accounted for 92.4% of referral traffic from standalone AI assistants, up from about 84% at the end of 2025. Gemini is a steady second. Claude grew 64x off a small base and passed Perplexity in March 2026, particularly among developers and technical buyers. Perplexity has fallen 61% from its early-2025 peak, partly because it increasingly keeps users inside its own tools instead of sending them out. Copilot has collapsed 96% from its high.
So among the assistants people open as separate apps, one platform does almost all the sending. That part of the finding is real and worth knowing.
the part the headline leaves out
Previsible excluded Google's AI Overviews and AI Mode from the study on purpose, because they don't produce a clean, trackable referral the way a standalone assistant does. And the report is direct about what that omission means: AI discovery happening inside Google almost certainly drives more traffic than every standalone assistant combined.
Flip the headline around and the real picture looks like this: ChatGPT wins the AI traffic that's easy to measure. Google, through the AI built into regular search, is very likely the bigger source of AI-influenced visits, it's just close to impossible to count cleanly with current tools. Treating 92% as a reason to ignore Google and chase ChatGPT alone means optimizing for the smaller, easier-to-see half of the problem.
the finding that should actually change your week
Here's the one that matters more than market share. Across platforms, roughly a quarter of AI-referred visitors get dropped onto a site's internal search results page instead of the page that actually answers their question. ChatGPT does this on nearly 29% of its referrals. The model trusts your domain enough to send someone your way, but can't always pick the exact page, so it hands the visitor your search box and lets them figure it out.
That turns internal site search from a housekeeping detail into an actual entry point. If a quarter of your hardest-earned AI visitors land on a weak search page and bounce, the model did its job and you lost the visitor at your own front door. Almost nobody treats this as a fixable problem, because almost nobody knows it's happening.
One more distinction worth knowing before you decide where to put your effort: ChatGPT and Gemini tend to trust a domain broadly and land visitors somewhere on it. Claude and Perplexity behave differently, they pick specific pages and lean toward long-form, well-sourced content. If your visibility strategy leans on in-depth editorial work rather than shallow pages, Claude and Perplexity can matter more than their small traffic share suggests.
what this means if you're a solo operator, not an enterprise SEO team
Most of the coverage of this study is written for marketing teams with a content calendar and a link-building budget. If that's not you, here's the version that applies:
- Don't panic-optimize for ChatGPT specifically. The 92% figure is about the traffic that's easy to measure, not the traffic that exists. Google's AI surfaces are the bigger prize and the one fewer people are chasing right now, which makes it the less crowded lane.
- Check whether AI crawlers can even reach your site before you write another word. Since mid-2025, a lot of hosting setups block GPTBot, ClaudeBot, and PerplexityBot by default at the firewall level, invisibly, with no error message and no traffic drop to notice. If that's you, none of the rest of this matters yet.
- Look at what happens after someone lands. If your own site search is weak, you're losing a real share of the AI traffic you do earn, at the door, after doing all the hard work of getting recommended in the first place.
- The underlying advice hasn't changed, and that's reassuring. Both the standalone-assistant story and the Google AI story reward the same thing: a site with real authority, specific sourced claims, and a clear, repeated answer to who you are and what you do. That's the same six-move foundation in how chatgpt decides what to recommend. Nothing here replaces that work. It adds two layers underneath it: can the model reach you, and where does it drop people once it does.
where to actually start
If you've never checked whether your site is reachable by AI crawlers at all, start there, it's a fifteen-minute check and the highest-leverage one available. If you're a creator wondering whether brands' AI tools can find you specifically, that's a different and equally practical question. Both are covered in this series, and both matter more than whichever platform currently owns the biggest headline.
run your first ai agent, free
see what ai actually surfaces about your business today. install your first working claude agent free, delivered on the page.
get your first agent freefrequently asked questions
Does ChatGPT really drive 92% of AI traffic?
It drives 92.4% of trackable referral traffic from standalone AI assistants, per Previsible's 2026 study of 6.77 million sessions across 166 sites. That figure excludes Google's AI Overviews and AI Mode entirely, which the study itself notes almost certainly represent more AI-influenced traffic than every standalone assistant combined.
Why are Google's AI Overviews excluded from AI traffic studies?
Because they do not generate a clean, trackable referral the way a standalone assistant does. The visit arrives through normal search infrastructure, which makes the AI influence real but close to impossible to isolate.
Should I optimize only for ChatGPT?
No. ChatGPT wins the traffic that is easy to measure. Google's AI surfaces are very likely the larger source overall and are less contested, because most operators assume AI SEO means optimizing for a chatbot.
What happened to Perplexity and Copilot?
Both declined sharply in the same dataset. Perplexity fell 61% from its early-2025 peak, partly because it increasingly keeps users inside its own tools, and Copilot collapsed 96% from its high.
What is the most useful finding in the study?
Not market share. Roughly a quarter of AI-referred visitors land on a site's internal search page rather than the page answering their question, and for ChatGPT it is closer to 29%. That is a fixable loss on traffic you already earned.
written by elisabeth hitz, certified in anthropic's ai fluency program (framework & foundations, and ai capabilities & limitations), plus claude 101 and claude cowork. sources: previsible, 2026 state of ai discovery report / third ai traffic study (6.77 million sessions, 166 sites, november 2024 to may 2026), as reported via search engine land, yahoo finance, and marketingtechnews. published 26 july 2026.