Traffic sent by ChatGPT to websites remains far lower than that from Google search, according to an analysis by Marie Haynes which concludes that Generative Engine Optimization, known as GEO, would be a misplaced investment for marketing teams. A separate study, covering 108 million sessions, reaches a similar conclusion, while noting that visitors coming from ChatGPT convert much better.
- ChatGPT sends far less traffic than Google search, under 1% of the volume.
- Visitors from ChatGPT convert at about 14%, against 3% for search.
- Analytics miss AI influence, because many choices are made inside the chat.
- Ignoring generative engines means losing presence at the moment of purchase.
Marie Haynes examined the SEO data of over 100 sites, totaling 145 million sessions, meaning visits recorded by analytics tools. She concludes that visits originating from ChatGPT represent less than 1% of the volume generated by traditional Google search.
The author of a follow-up study analyzed 108 million sessions, recorded from January 1 to November 2, 2026, across more than 57 sites. AI-generated traffic there represents 0.66% of the total, meaning, for every 1,000 users coming from search engines, about seven users are sent by AI, provided they are identifiable by measurement tools. This ratio varies significantly by site, some reaching 20 to 25 AI users per 1,000, but it remains low overall.
The same type of data, however, shows a significant conversion gap. According to Marie Haynes’ study, visitors arriving from ChatGPT convert at nearly 14%, compared to 3% for visitors from traditional search. The data thus indicates a conversion rate approximately five times higher, even though, according to the same analysis, Google remains responsible for the vast majority of total sales.
For the author, the reasoning is based on an inadequate measurement. Recorded sessions count no value when a generative engine satisfies a query without sending the user to an external site. Yet, when a purchasing team compares software or a buyer researches touring bikes, the bulk of the comparison can take place within the conversation itself. The user then visits the brand’s site by typing its name into a traditional search engine, and the analytics attributes the sale to organic search, whereas the choice was actually made in the chat interface.
A value in traffic quality
According to the author, a visitor who consults a generative answer and then follows a link to a site demonstrates strong commercial intent, as they have already moved beyond basic information gathering. Conversely, traditional search traffic includes accidental clicks and curiosity visits, which inflate volumes without generating sales. For B2B suppliers, specialized consultants, or niche retailers, the total number of visitors is therefore, in his view, a misleading metric.
The author believes the logical response to such a conversion advantage is to study the reasons for this success and ensure the company appears when serious buyers ask for recommendations. He also notes that generative engines take into account relationships between entities and structured knowledge, rather than just traditional ranking signals. A company poorly represented during this silent research phase therefore risks losing a conversion that might later have come through traditional search.
A change in behavior
The author finally notes that autonomous systems will alter commercial activities on digital channels. In his view, ignoring generative engines amounts to forfeiting presence at the moment of purchase decision, to the benefit of competitors who measure their influence.
