According to OpenAI's estimate, the volume of search queries in ChatGPT has reached approximately 20% of Google's volume, meaning the search market leader processes about five times more. The statement relies on the company's own definition of 'search' and has already sparked debate: the internal volume of assistant queries and its actual impact on web traffic are two different metrics, and the difference between them is key to understanding the news.

image
image

What happened

The statement was made by OpenAI's head of consumer products, Adam Fry: according to him, the volume of queries triggering a web search within ChatGPT is about 20% of Google's search volume, meaning Google processes approximately five times more. Fry spoke on November 2, 2026, at a dinner organized by OpenAI in New York, and repeated this in an interview with Alex Kantrowitz of Big Technology. The specific methodology formulation is significant: OpenAI defines 'search' as any user query that led to the initiation of a web search within ChatGPT. Another scale metric was also mentioned at the time: according to company data, approximately 250 million people talk to ChatGPT, a figure highlighted in the headline of the Big Technology episode.

Context

The statement is easier to understand if two fundamentally different metrics are separated. The first is the internal volume of queries within the assistant: ChatGPT works as a search pipeline where a user's question is followed by the initiation of a web search and the synthesis of an answer, with the decision to connect a web search made by an internal triggering policy that depends on the product version and can be revised. The second metric is the external effect on websites, which is measured by Ahrefs: according to the service's data for October 2025, Google provides 24.74% of referral traffic to websites, while ChatGPT provides only 0.36%. These numbers do not contradict each other but highlight a gap: activity within the assistant currently barely converts into link clicks. The timing of the statement is also telling: by autumn 2026, competition between OpenAI and Google is increasingly measured by product volumes, i.e., the number of queries and audience size, rather than model benchmarks, so figures like the ratio to Google are becoming a new currency for public comparisons.

Why this matters for the industry

For the industry, the statement is more of a narrative and negotiation signal than a technical event: it does not change models, benchmarks, or architectures, but it sets the framework for discussion and will likely become a standard slide in AI startup pitches and an argument in debates about the fate of SEO. Two practical consequences are important. Business models tied to traffic from classic search are not yet under pressure from ChatGPT: the referral traffic channel from the assistant is negligible, and it is too early to shift budgets and analytics to it. At the same time, a window is open for builders on top of assistant search: prototypes of 'question with web search plus action completion' can already be built using existing ChatGPT search tools and agent connectors, and the layer of brand presence in AI answers is largely unoccupied, so companies can cheaply test today how assistants describe their products and correct descriptions, documentation, and structured data. A separate rule for product teams: a self-declaration by an interested party with a convenient metric definition is not reproducible evidence and cannot be relied upon for product decisions.

Why this matters for users

It is most useful for readers not to conflate 'search in ChatGPT' and 'traffic to websites,' because they are different things. If you ask ChatGPT a question with web search enabled, the answer is synthesized within the chat, and websites receive almost no visitors from this: sources that readers are used to via search links fit poorly into the assistant model. Google remains the main channel that brings people to websites, and assistants currently lack a familiar equivalent of a link list where the choice of sources remains with the reader. Using ChatGPT instead of a search bar is already participation in the shift of the entry point to information; however, OpenAI's figure does not mean that the habits of most readers have changed: it describes the volume of queries within one product, not the share of the search market.

What is currently unknown / limitations

OpenAI's statement lacks absolute numbers, measurement windows, deduplication rules, and aggregation methods, making the comparison non-reproducible and remaining a self-declaration by an interested party. The definition of 'search' depends on the internal web search triggering policy, which may change between ChatGPT versions, meaning the metric is unstable over time and not directly comparable to Google search queries. Ahrefs data relates to October 2025 and measures a different metric, referral traffic, so current query volume or its dynamics cannot be derived from it. Finally, only one point value is available: it is impossible to judge the trend, the speed of the shift, or whether independent tools like panel data and third-party traffic analytics will confirm the figure from it.

Sources

Author

Look at AI, editorial team