News & analysis
OpenAI: ChatGPT Search Volume Is About 1/5 of Google's
OpenAI's Adam Fry says ChatGPT handles roughly one-fifth as many search-like queries as Google, a figure built on a different counting method than outside trackers use.
By Roomer Editorial · · 3 min read
What did OpenAI actually claim?
ChatGPT consumer product lead Adam Fry said ChatGPT's search volume is roughly one-fifth of Google's. He made the comment during an open discussion with Alex Kantrowitz at a Monday dinner hosted by OpenAI in New York. Big Technology posted the transcript Tuesday.
Fry defined his terms narrowly. OpenAI counts a search every time a user's question sends ChatGPT out to the web. Asked directly whether a full conversation counts as one search or each web-querying question counts separately, Fry said:
He said early ChatGPT use skewed toward writing and coding help, the kind of task people wouldn't normally search for. Over the past year, he said, that's shifted toward questions about people, places, things, finance, and health, where the product reaches out to the web for current information.
Fry did not say which Google number he was comparing against, what time period the ratio covers, or which regions it includes. He also didn't answer when asked whether being one-fifth of Google's size would make ChatGPT several times larger than Bing.
How does this compare to other estimates?
Big Technology's report notes ChatGPT's weekly active users grew 20 percent, from 1 billion to 1.2 billion, over the prior two months. That growth is the context for Fry's comment, not a verification of it.
Ahrefs ran its own estimate in February and landed at roughly 12 percent, not 20. Ahrefs started from a Harvard University and OpenAI study that put ChatGPT at 2.5 billion daily prompts in July 2025, then found about 65 percent of those resembled the kind of query people used to type into Google. That share was compared against Google's own disclosure of more than 5 trillion searches per year, announced in March 2025.
The two estimates don't measure the same thing. Ahrefs sorts prompts by request type to decide what counts as "search." Fry's number counts only the subset of questions that trigger an actual web lookup inside ChatGPT. Different filters, different outputs.
Ahrefs also runs a separate referral tracker across sites using its free Web Analytics tool. In October, across 118,671 tracked sites, Google sent 24.74 percent of traffic and ChatGPT sent 0.36 percent. That's the one dataset here that measures something concrete and external: actual visits landing on actual websites, not self-reported query counts.
Does any of this show ChatGPT replacing Google search?
No, and the source is explicit about that gap. Neither Fry's one-fifth figure nor Google's own growth numbers say whether ChatGPT queries are substituting for Google searches or simply adding new usage on top.
On Google's side, Alphabet's Q3 release had CEO Sundar Pichai saying the company's year-over-year query growth rate is the highest it's been in five years, with AI investment cited as a driver. That's a growth rate, not a volume, and it doesn't isolate ChatGPT's effect on it.
Put the three datapoints side by side: Fry's query-count ratio (~1:5), Ahrefs' prompt-based ratio (~1:8), and Ahrefs' referral-traffic ratio (~1:69). They're not contradictory so much as they're answering different questions — how often people ask, versus how often that produces a click to a website.
Roomer Editorial's take
The number that matters to a travel creator isn't query volume. It's referral traffic — whether a reader ends up on your site after asking an AI tool a question. On that measure, the only external data in this story (Ahrefs' tracker) shows Google still sending roughly 69 times more visits than ChatGPT, even as ChatGPT's own usage numbers climb.
That gap is why we'd keep building distribution the boring way: your own site, your own published pages, content that shows up in Google's index where the traffic still overwhelmingly is. Chasing AI-referral traffic as a current strategy means optimizing for a channel that, per this data, sends well under one percent of the volume Google does.
The exception is worth watching, not acting on yet. If OpenAI's weekly actives keep growing 20 percent every couple of months, the referral math could shift fast enough to matter within a year. Until Ahrefs' monthly tracker shows that shift showing up in actual site visits, don't restructure a content plan around it.
One action: check your own site's monthly referral mix and see what share, if any, comes from ChatGPT right now — it's probably close to Ahrefs' 0.36 percent average, and that's your honest baseline before betting on anything changing.
Source: Search Engine Journal
