ChatGPT Search Volume Is About One-Fifth of Google's

ChatGPT Search Volume Is About One-Fifth of Google's

By · November 4, 2026 · 9 min read


OpenAI says ChatGPT gets about one-fifth as many search-like queries as Google. See how it counts a search and what outside data shows.

What OpenAI said about ChatGPT search volume

ChatGPT receives roughly one-fifth as many search-like queries as Google, according to Adam Fry, the consumer product lead for ChatGPT. The comparison comes from OpenAI's own analysis, which put the share at about 20%. That makes it different from the outside estimates that have circulated until now, because it was offered by the company that runs ChatGPT.

Fry made the point in an open discussion with journalist Alex Kantrowitz at a Monday evening dinner in New York hosted by OpenAI. A transcript was published the next day.

There is a catch that matters for anyone planning around the number. It measures search-like queries. It does not measure how many people ChatGPT sends to websites, which is a separate metric, and outside data on that traffic puts ChatGPT far behind Google.

How OpenAI counts a search

Kantrowitz asked Fry a practical question: if a conversation includes a search, does it count as one search, or does each question that needs a web search count separately? Fry answered: "Every time you ask a question and go search the web, that's a search."

So the unit is the trip to the web. A long back-and-forth conversation does not count every message as a search. Only the questions that send ChatGPT out to look something up are counted, and each of those is counted on its own.

That definition is narrower than "everything people type into ChatGPT." Most prompts, such as a request to rewrite an email or fix a function, never touch the web, and they do not appear in this figure.

How people's use of ChatGPT has changed

Fry said people first came to ChatGPT mostly for writing and coding help, tasks they would not have run through a search engine. Early versions also had no up-to-date information. Users would sometimes get a reply saying the question fell after the model's training date.

That has shifted over the past year. People now bring questions about people, places and things that need current information, along with finance and health questions where an accurate, recent answer matters. Those are the cases where ChatGPT does a web search. Fry described the category as anything users would have taken to a traditional search engine.

He also pointed to personalization as a reason people choose ChatGPT for this. A search box handles a one-off question about the Eiffel Tower. ChatGPT can answer the same question while taking into account that the person is planning a trip to Paris, because it builds up context over time.

What Fry did not say

The published exchange leaves several things open. Fry did not say which Google figure the comparison used, what period it covers or which regions it includes. Without those three details, the one-fifth ratio is hard to line up against any other estimate.

Kantrowitz also asked whether a figure of one-fifth of Google would make ChatGPT several times larger than Bing, which holds roughly 5% of the search market. Fry did not address the comparison and moved to talking about ChatGPT's user growth instead.

ChatGPT user growth

On growth, Fry cited weekly active users rising from 1 billion to 1.2 billion over the last two months, an increase of about 20%. He described that as roughly 100 million new users every 30 days, and said the company stays focused on how fast ChatGPT is growing.

Growth in users does not translate directly into growth in searches, since a new user may use ChatGPT only for writing or coding. Still, it gives context for why OpenAI is willing to talk about search at all. A larger base of weekly users means more chances for questions that need the web.

What outside data shows

Fry's number is not the only estimate. In February, the SEO software company Ahrefs estimated that ChatGPT's search volume was around 12% of Google's. It used a different definition of search from the one OpenAI uses, and the two results should not be treated as competing answers to the same question.

Ahrefs started with 2.5 billion daily ChatGPT prompts in July 2025, a figure taken from a study by Harvard University and OpenAI. It then determined that about 65% of those prompts were the kind of query people would previously have looked up on Google. Finally, it compared the result with the more than 5 trillion searches a year that Google announced in March 2025. That arithmetic gives roughly 12%.

The author of the Ahrefs analysis acknowledged that his 65% was higher than the share the researchers themselves had classified as search. The study he drew on sorted prompts by the type of request, and Ahrefs applied its own judgment about which of those would once have been Google searches. Fry's figure works differently, because it counts web searches that ChatGPT actually runs. One method looks at what kind of request a prompt is. The other counts what ChatGPT did.

That difference helps explain why a 12% estimate and a 20% claim can both exist without either one being wrong. They answer slightly different questions, using different inputs, and neither has been checked against the other.

ChatGPT and Google referral traffic

Volume of queries is one thing. Traffic sent to websites is another, and here the gap is wide.

Ahrefs also tracks where website traffic comes from across sites that use its free Web Analytics tool. In October, across 118,671 sites, Google accounted for 24.74% of tracked traffic and ChatGPT for 0.36%, according to the tracker's November 3 update. By simple division, that is about 69 visits from Google for every one from ChatGPT in that sample.

The sample has limits. It covers only sites that use one company's free analytics tool, so it is not a count of the whole web. It is still a useful signal, because it measures something that can be checked in site logs: people who arrived at a page.

Put the two measures side by side and an odd picture appears. If ChatGPT handles one query for every five that Google handles, but sends a small fraction of the visits, then many ChatGPT searches end inside ChatGPT. The person gets an answer and does not click through. The data does not prove that, and it would take click studies to confirm it, but it fits the numbers.

What Google reports

Google does not publish a search count that can be set against Fry's figure. Alphabet's third-quarter release reports query growth as a rate, not as a total.

In that release, CEO Sundar Pichai said AI investments are driving overall search query growth. He said Google's "year-over-year query growth rate is the highest it's been in the last five years."

That statement is relevant because it cuts against a simple story in which ChatGPT takes searches away from Google. If Google's own query growth is at a five-year high, then total searching may be expanding, with ChatGPT adding to it, rather than splitting a fixed pool. Neither Fry's number nor Google's growth rate settles the question, since neither shows where ChatGPT's searches came from.

Query volume and referral traffic are different metrics

The whole story comes down to keeping three numbers apart: OpenAI's figure counts search-like queries, defined as questions that send ChatGPT to the web. The Ahrefs referral tracker counts visitors who arrive at websites from each source. Google's growth rate describes change in queries over a year and gives no count at all.

Reading any one of these as if it were another leads to the wrong conclusion. A team that sees "one-fifth of Google" and assumes ChatGPT will send one-fifth of the traffic will be badly off. A team that sees 0.36% of referral traffic and assumes ChatGPT is a minor way of finding information will undercount how much searching happens inside the product.

What marketers and publishers can take from this

When you report on search, say which metric you are using. If a channel plan includes ChatGPT, label whether the number is query volume or referral traffic, since they tell different stories.Do not assume ChatGPT searches are replacing Google searches. Nothing published so far shows that.Track ChatGPT referrals separately in your analytics so you can see the real traffic for your own site, instead of relying on an industry-wide estimate.Treat OpenAI's ratio as an approximate figure until the company says which Google number, period and regions it used.

What is still unknown

Several important details are missing. OpenAI has not said which Google figure it compared against, what period the comparison covers or which regions are included. Fry did not answer the question about Bing. And no one has shown whether a search done in ChatGPT is a search that would otherwise have gone to Google, or one that would never have been run at all.

What is known is the ratio OpenAI gave, the rule it used to count a search, and the gap in referral traffic that outside data shows. Any claim that goes beyond those three points rests on assumptions.

Frequently asked questions

How does ChatGPT's search volume compare with Google's?

According to Adam Fry, ChatGPT's consumer product lead, ChatGPT receives roughly one-fifth as many search-like queries as Google. OpenAI's analysis put the share at about 20%.

How does OpenAI count a search in ChatGPT?

Fry said every time a user asks a question and ChatGPT goes to the web, that counts as a search. A conversation does not count each message automatically.

Where did Fry make these comments?

He spoke in an open discussion with Alex Kantrowitz at a Monday evening dinner in New York hosted by OpenAI. A transcript was published the next day.

What did Fry leave out?

He did not say which Google figure the comparison used, what period it covers or which regions are included. He also did not answer a question about how ChatGPT compares with Bing.

How many weekly users does ChatGPT have?

Fry cited growth from 1 billion to 1.2 billion weekly active users over the last two months, an increase of about 20%.

Is the one-fifth figure the same as Ahrefs' estimate?

No. In February, Ahrefs estimated ChatGPT's search volume at around 12% of Google's, using a different definition of search. It classified about 65% of ChatGPT prompts as the kind users would once have searched on Google, while OpenAI counts the web searches ChatGPT runs.

How much website traffic does ChatGPT send compared with Google?

In October, across 118,671 sites in Ahrefs' tracker, Google accounted for 24.74% of tracked traffic and ChatGPT for 0.36%, according to the November 3 update.

Is ChatGPT replacing Google search?

The available data does not show that. Neither Fry's figure nor Google's year-over-year query growth rate says whether ChatGPT searches replace searches on Google. Sundar Pichai said Google's query growth rate is the highest it has been in five years.

What should marketers do with these numbers?

Report query volume and referral traffic as two separate metrics, state the definition behind each, and track ChatGPT referrals on your own site instead of relying on industry averages.

Reference

Search Engine Journal: https://www.searchenginejournal.com/googles-search-volume-is-about-5x-chatgpts-openai-says/595817/

Ahrefs (AI vs search benchmarks): https://ahrefs.com/benchmarks/ai-vs-search

Big Technology: https://www.bigtechnology.com/p/openai-chatgpt-search-is-about-20

Ahrefs (ChatGPT search volume study): https://ahrefs.com/blog/chatgpt-has-12-percent-of-googles-search-volume/

Tags: ChatGPT search volume, ChatGPT vs Google, Google search volume, OpenAI, ChatGPT search, search-like queries, AI search traffic, CHAT-GPT weekly active users, referral traffic