OpenAI puts ChatGPT searches at one-fifth of Google’s

OpenAI puts ChatGPT searches at one-fifth of Google’s

ChatGPT reaches nearly 20% of Google’s search volume, according to Adam Fry, OpenAI’s head of consumer product for the chatbot. The comparison points to a shift in how people search for information, but it does not mean ChatGPT sends websites one-fifth of the traffic they receive from Google. These are different metrics, and that distinction matters to any business that depends on search to attract visitors.

Fry shared the figure in a conversation with Alex Kantrowitz of Big Technology during a dinner hosted by OpenAI in New York. The counting rule he described is specific: every question that prompts ChatGPT to search the web counts as a search, even if it is part of a multi-turn conversation.

“Every time you ask a question and go search the web, that’s a search.”

— Adam Fry, head of consumer product for ChatGPT

A query is not the same as a visit

The 20% compares search queries, not clicks or users. ChatGPT can answer within the conversation; a question triggering a web search does not mean the user will then visit an external page. Mixing the two measures would turn a comparison of activity into a different claim about the distribution of visits.

Ahrefs data shows the gap between these two indicators. In October, its tracking of 118,671 sites recorded Google as the source of 24.74% of measured traffic and ChatGPT as the source of 0.36%, according to an update published on November 3. This is a sample of sites using Ahrefs’ free Web Analytics tool, not a measurement of the entire internet; even so, it makes clear that the query volume cited by OpenAI does not automatically translate into a similar share of referrals to pages.

There are also differences between methods for estimating how many searches ChatGPT represents. In February, Ahrefs estimated a figure close to 12% of Google’s volume. It started with an estimate of 2.5 billion ChatGPT prompts per day in July 2025 and classified around 65% of them as searches. That approach classified the type of request; OpenAI, by contrast, counts questions that trigger a web search. That is why the 12% and 20% figures are not directly comparable.

What OpenAI counts—and what it leaves unspecified

OpenAI’s definition excludes tasks that do not require up-to-date information, such as much of the writing and coding help that Fry says people initially used ChatGPT for. The company says that over the past year users have also brought the chatbot questions about people, places, and things, as well as financial and health queries, which may require a web search.

But details needed to interpret the ratio remain unclear: Fry did not identify which Google figure he used as a reference, the comparison period, or the regions included. Google said in March 2025 that it handled more than five trillion searches annually, but OpenAI did not say that this was the basis for its calculation. Without that denominator and the time period, the 20% serves as a rough reference, not a series that can be reproduced or compared precisely.

ChatGPT’s scale helps explain why OpenAI presents the figure as significant. Big Technology reported that the service grew from 1 billion to 1.2 billion active users in two months. There is, however, a discrepancy in how that activity is described: Big Technology refers to weekly active users, while Punto Informatico refers to monthly active accounts. Both publications report the same growth figure, but not the same frequency.

Google and OpenAI are also competing over the answer

The competition is not just about adding up queries. Google is expanding its artificial intelligence features in Search and AI Mode; in addition, its CEO, Sundar Pichai, said that year-over-year growth in Google search queries was the highest in five years, attributing some of the momentum to AI investments, according to Search Engine Journal. The figure does not provide an absolute volume to compare with OpenAI’s, but it does indicate that Google does not describe its search engine as a channel in decline.

For OpenAI, the bet is that some questions people once typed into a search engine will now be asked in a chat that can search the web and build an answer. Fry also described a more interactive direction: “You’re talking to a canvas”, he told The Verge, discussing visual interfaces within the chatbot. If an answer brings together an explanation, links, and interactive elements on a single screen, the product’s value is measured not only by how many queries it processes, but by how much of the task it completes without sending the user to another page.

This shift creates a tension for companies that publish content. A query answered within ChatGPT can give users a quick response, but it does not necessarily bring them to the site that provided the information. For those measuring performance by referral traffic, Ahrefs’ figure is closer to the immediate business question than OpenAI’s 20%. For those planning their search presence, however, the query figure may signal that a growing share of questions with search intent is being asked in a conversational environment. These are related signals, not interchangeable ones.

Advertising and measurement: two business fronts

Monetization adds another difference between the services. The Verge reported that OpenAI announced it would begin introducing sponsored images of products and services in ChatGPT, after showing names, logos, and links. Fry said advertising had not harmed the product or its usage, adding that most users did not seem to react negatively. He also suggested that OpenAI could show more ads. Google, meanwhile, faces a tension acknowledged by 9to5Google: the company admits that monetization is lower as it expands AI features in Search.

For companies that buy advertising or depend on organic visits, it is not enough to add “AI searches” to a report and compare them with Google queries. OpenAI’s reported volume counts questions that trigger web searches; Ahrefs’ tracking measures visits to sites. Separating the two indicators makes it possible to evaluate whether a presence on ChatGPT generates exposure, referrals, or conversions, without assuming that an increase in queries automatically produces more traffic.

OpenAI presents a chatbot that already processes a considerable fraction of Google’s query volume, but its percentage depends on its own definition and a comparison baseline that Fry did not detail. The next question for businesses is not just how many searches are made on ChatGPT, but how many end up bringing users to their pages.

Sources: Punto Informatico, Search Engine Journal, 9to5Google, Big Technology, The Verge