ChatGPT Sends Just 0.62% of Google's Organic Traffic, Study Finds

Across 100 websites and more than 145 million GA4 sessions over 12 months, ChatGPT drove just 0.62% of the session volume that Google Organic delivered. Add every other large language model and the figure reaches only 0.73%. That is the central finding of a new analysis by search consultant Marie Haynes, who argues that the industry's obsession with GEO — generative engine optimization, the practice of tuning content to be cited by AI tools — is aimed at the wrong target.

Her data is blunt. For the vast majority of sites, LLM traffic is minuscule next to Google organic. And the growth story many predicted has already reversed.

Do LLM visitors convert better?

Partly, yes — and the numbers are striking. Of the 100 sites Haynes studied, 55 tracked conversions in GA4:

• Google organic visitors converted at 3.02%.
• ChatGPT visitors converted at 13.8% — more than 4.5 times the rate.

Scale erases the advantage. Google still drove 96.5% of all conversions across the channels analyzed, and Google organic produced 27 times more conversions than every LLM combined.

Did the exponential LLM traffic trend hold?

No. When ChatGPT launched, many practitioners watched referral clicks climb and assumed LLMs would eventually out-send Google. Haynes says they were wrong. Using Google's Antigravity tool, she calculated collective ChatGPT visits across the 100 properties: in August 2025, coinciding with the release of GPT-5, the exponential growth ended. Since then, ChatGPT has been sending fewer and fewer visitors to websites.

What does this change for SEOs?

The AI search actually sending people to businesses today is not ChatGPT, Haynes argues — it is Google itself. Between AI Overviews, AI Mode, and a core ranking system now largely driven by machine learning, millions of people use AI-assisted search daily without leaving Google.

She expresses frustration at how much money GEO-tracking companies have raised to monitor brand mentions in ChatGPT. "We are chasing the new thing," she writes.

Her practical advice is measurement-first. Rank tracking means little, and Search Console offers little that is immediately actionable, so businesses should define what a conversion or key event looks like and track it. In her sample, 45 of the 100 sites were not tracking any key events at all. Her broader recommendation: focus on improving user satisfaction far more than optimizing for machines, because Google's ranking systems continually work to predict what searchers will find most helpful.

Where the real opportunity lies: agentic search

Google CEO Sundar Pichai recently said: "Search will become an agent manager" — shifting search from a passive way to find links to an active work coordinator. At Google I/O, Sergey Brin said of the web in ten years: "I don't think we really know what the world looks like in 10 years." Demis Hassabis added: "The web is going to change quite a lot if you think about an agent-first web… So I think things will be pretty different in a few years."

Haynes identifies four concrete developments to watch:

• Universal Cart. Google's system aggregates products from multiple retailers into a unified checkout inside Google Wallet, monitoring prices, promotions and restocks. Shoppers can buy across vendors without visiting a single site. Opportunity: master Merchant Center, which now includes a helpful AI assistant, so products get bundled into Universal Carts.
• Universal Commerce Protocol (UCP). Governed by Google, Shopify and Stripe, UCP is an open discovery layer letting AI agents compare inventory and purchase natively while the seller remains merchant of record. Opportunity: technical SEO expands to agentic discovery endpoints such as /.well-known/ucp, plus deeper competitor research — harder now, because no two Google searches produce identical results.
• WebMCP (Model Context Protocol). Websites can expose interactive functions — forms, calculators, booking schedulers — directly to AI agents via declarative markup and APIs. Haynes calls it early but says experimenting now beats waiting, and suggests brainstorming with AI on what tools agents would want to use.
• Intelligent Eyewear. Google's preferred term for its Gemini-powered Android XR devices: audio frames with Warby Parker and Gentle Monster first, then micro-LED display glasses, arriving this fall. Haynes believes they will expand rather than replace search — think guided cooking subscriptions or "Hey Gemini, find me that shirt and buy it."

The takeaway

Chasing ChatGPT citations optimizes for the wrong target, Haynes concludes. Businesses should study how Google Search and its AI surfaces represent their brand, and invest time preparing for the agentic web through UCP, WebMCP and intelligent eyewear. She also invites practitioners to publish their own measurement work as search shifts — and includes a ready-made prompt for readers to identify agentic commerce opportunities for their own business. The next signal to watch: whether ChatGPT referrals keep declining and whether Google's agent-driven commerce surfaces reach users at scale.