At an OpenAI press event on Nov. 2, product lead Adam Fry said ChatGPT’s search volume was roughly one-fifth of Google’s query volume, according to a report published by 9to5Google on Nov. 3. Fry’s comparison counts a ChatGPT search when a prompt triggers a web search. It does not specify a shared counting method or Google’s query total, so the figure is an attributed comparison, not a like-for-like measure of search-market share.
A search count and a citation share answer different questions: one tracks search activity, the other tracks which sources an AI answer cites.
What the reported search-volume figure counts
The ChatGPT measure described by Fry starts with a prompt that leads to a web search. Google’s side of the comparison is described as query volume. Without a common counting method and a stated Google total, the roughly 20% figure cannot be read as a comparable share of Google searches.
What GrackerAI measured in its SIEM study
GrackerAI’s October 2026 benchmark examined enterprise security information and event management (SIEM) searches. Its report describes 10,000 buyer-intent queries and 160,000 responses across four engines: ChatGPT, Gemini, Google AI Overview and Google AI Mode. The study covered the United States, Canada, India and Germany, with 40,000 responses per engine, and counted 954,000 cited sources. GrackerAI’s SIEM benchmark
A Security Boulevard report put the broader GrackerAI benchmark series at 1.74 million AI answers across eight studies. That series-wide total is separate from the 160,000 responses in the SIEM study.
Vendor-site citations varied by search surface
In the SIEM benchmark, GrackerAI reported that vendor websites accounted for 94.9% of ChatGPT citations and 30.5% of Google AI Overview citations.
| Search surface | Share of citations from vendor websites |
| ChatGPT | 94.9% |
| Google AI Overview | 30.5% |
Across the four engines, the highest pairwise Jaccard similarity for cited domains was 0.29. Jaccard similarity measures how much two sets overlap; here, the sets were the domains cited by each engine. The result points to distinct citation patterns among the four engines for the SIEM queries in this study. For this category, a citation pattern on one search surface cannot be assumed to match another’s.
