AIThis post was created with the assistance of artificial intelligence (AI).

By Thorsten Meyer | November 5, 2026 | ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

On Monday evening, at a dinner OpenAI hosted in New York, the man in charge of ChatGPT’s consumer product made a big claim. “ChatGPT’s volume of search-like queries is now roughly one-fifth the volume of queries of Google,” Adam Fry said.

One fifth of Google. For context: Bing, after more than fifteen years, gets less than 5 percent of the clicks that search engines send to websites, according to StatCounter.

So is ChatGPT really catching up with Google? We put OpenAI’s number next to every independent yardstick we could check: Google’s own figures, OpenAI’s own published usage data, two modeled estimates from search companies, and two traffic panels. The short answer: the number may be true by OpenAI’s own definition, but nobody outside OpenAI can check it, and the independent data tells a very different story about what reaches websites.

Key takeaways

  • The claim: OpenAI’s Adam Fry says ChatGPT handles “roughly one-fifth” of Google’s search-like query volume. He said it on 2 November 2026 in New York; Big Technology published the transcript on 3 November.
  • The definition: “Every time you ask a question and go search the web, that’s a search.” OpenAI did not say which Google figure it compared against, for which period or which countries.
  • Independent estimates are lower. Ahrefs estimated about 12 percent in February. seoClarity estimated about 7.5 percent on 28 October. Both are models, not measurements.
  • OpenAI’s own data raises a question. In OpenAI’s published usage data, “seeking information” was 20.2 percent of consumer ChatGPT messages in September 2026. For one fifth of Google, by our rough arithmetic, about six in ten of all messages would have to trigger a web search.
  • Clicks are a different world. In Ahrefs’ panel of 118,671 websites, Google sent about 69 times as much traffic as ChatGPT in October.
  • Google is still growing. Alphabet says its query growth in the third quarter was the fastest in five years, and search revenue rose 12 percent to $63.35 billion.
  • For companies: being cited in AI answers now matters, but it does not replace search traffic yet. Measure both.
  • For users: an answer with sources is only as good as the sources. Click at least one when it matters.
Video: OpenAI’s one-fifth search claim checked against Google’s figures, OpenAI’s own usage data and independent traffic panels, with animated charts, in about five minutes. Narration: AI voice clone of Thorsten Meyer; images AI-generated.

Products Worth Considering

LG UltraGear 34GX90SA-W 34-inch WQHD Curved OLED Gaming Monitor

LG UltraGear 34GX90SA-W 34-inch WQHD Curved OLED Gaming Monitor

As an affiliate, we earn on qualifying purchases.

What OpenAI actually said

The setting matters. This was not a blog post, a filing or a published study. It was a short fireside chat between Adam Fry, who runs ChatGPT’s consumer product, and the journalist Alex Kantrowitz at an OpenAI-hosted dinner in New York on Monday, 2 November. The Verge, which was there, calls it a press event. Kantrowitz’s newsletter Big Technology published the conversation the next day.

Fry gave three things:

  • The ratio. “ChatGPT’s volume of search-like queries is now roughly one-fifth the volume of queries of Google.”
  • The definition. Asked whether a whole conversation counts as one search or every question counts, Fry said, as quoted by Search Engine Journal and Search Engine Roundtable: “Every time you ask a question and go search the web, that’s a search.” The Verge adds that defining a word or a simple how-to explanation does not count.
  • The growth. “ChatGPT has gone from one billion to 1.2 in just the last two months,” he said about weekly users.

What he did not give: which Google number OpenAI divided by, the period, the countries, or any underlying data. The Verge notes that Fry “did not specify the source of the Google search volume number.” Google did not comment on the claim in any of the reports we read.

So the claim is confirmed as a statement. It is not confirmed as a fact. That is the gap this article tries to measure.

Five yardsticks, five different answers

When people compare ChatGPT with Google, they often mean different things. Some count questions. Some count prompts that look like searches. Some count the clicks that land on websites. Here is what each yardstick says.

Thorsten Meyer AI
Five yardsticks

Infographic 01 · ChatGPT as a share of Google

Questions or clicks? ChatGPT is a fifth of Google, or less than 2 percent

OpenAI says one fifth. Independent models say less. Traffic panels say far less, because a chat answer often ends without a visit.

Counting questionsHow many search-like questions?

  • OpenAI (Adam Fry)search-like questions that trigger a web lookup · Google baseline not named≈ 20 %
  • Ahrefs modelprompts of a kind people used to search on Google≈ 12 %
  • seoClarity modelseeking information + a quarter of practical guidance≈ 7.5 %

Counting clicksHow many visits reach websites?

  • Ahrefs traffic panelvisits sent to 118,671 websites, Oct 2026≈ 1.5 %
  • StatCounterclicks to sites in its network, monthly range0.08–0.41 %

All values: ChatGPT as a percentage of Google on the same yardstick. OpenAI’s figure is a company claim (striped bar); the two models are estimates; the panels see only their own sites. Bars scaled to 20 %.

Sources: Big Technology, 3 Nov 2026 · Ahrefs, 11 Feb 2026 · seoClarity, 28 Oct 2026 · Ahrefs traffic panel, Oct 2026 · StatCounter, Oct 2025–Oct 2026.
Powered by Thorsten Meyer AI

  • OpenAI (2 November 2026): about 20 percent. Search-like questions in ChatGPT that trigger a web lookup, compared with an unnamed Google figure.
  • Ahrefs (11 February 2026): about 12 percent. Ahrefs’ Patrick Stox took 2.5 billion ChatGPT prompts per day from OpenAI’s 2025 usage study and counted 65 percent of them as things people used to search on Google. He says himself that the 65 percent is his own judgment and “can be lower depending on how you define search.”
  • seoClarity (28 October 2026): about 7.5 percent. seoClarity extrapolated ChatGPT’s message volume with OpenAI’s user growth, counted “seeking information” plus a quarter of “practical guidance” as search, and compared that with Google’s five trillion yearly searches plus an assumed 20 percent growth. Its own words: “a modeled comparison, not a measured market-share figure.”
  • Ahrefs traffic panel (October 2026): about 1.5 percent. Across 118,671 websites using Ahrefs’ analytics, Google brought 24.74 percent of all visits and ChatGPT 0.36 percent.
  • StatCounter (October 2025 to October 2026): under half a percent. StatCounter counts clicks from a search engine or chatbot to the more than one million sites in its network, not searches. In its Google-versus-ChatGPT comparison, ChatGPT’s share of those clicks ranged from 0.08 to 0.41 percent per month.

None of these numbers is wrong. They measure different things. The first three ask how often people use ChatGPT like a search engine. The last two ask how often that use sends someone to a website. That difference is the real story for anyone who runs a website or a business.

A reality check with OpenAI’s own data

OpenAI publishes more than talking points. Its economic research team released a dataset called Signals on 26 October 2026, built from a sample of about one million consumer ChatGPT messages per month. It covers Free, Go, Plus and Pro accounts and leaves out Enterprise and Codex.

In that data, “seeking information” grew from 16.7 percent of messages in July 2024 to 20.7 percent in August 2026, and stood at 20.2 percent in September. Writing fell from 34.1 to 21.4 percent over the same time. People really do use ChatGPT more like a search engine than two years ago. OpenAI’s own numbers show that trend.

Thorsten Meyer AI
Reality check

Infographic 02 · OpenAI’s claim against OpenAI’s data

More people use ChatGPT like a search engine. One fifth of Google is still a stretch

Looking things up has nearly caught up with writing in ChatGPT. The arithmetic below is ours, with every assumption stated.

What people use ChatGPT for

Share of consumer ChatGPT messages per month (Free, Go, Plus, Pro; without Enterprise and Codex)

10%20%30%40%Jul 2024Jul 2025Sep 202621.4%20.2%34.1%16.7%
Seeking informationWriting

Our rough arithmetic

  1. Google, per day (at least)13.7 bn“More than 5 trillion searches” a year, January 2025 data (Google). One fifth: at least 2.7 bn.
  2. ChatGPT messages per day, July 2025≈ 2.6 bn18 billion a week, 700 million weekly users (OpenAI usage study).
  3. Scaled to 1.2 bn weekly users≈ 4.4 bnAssumes each user still sends as many messages as in 2025 (our assumption).
62%The number to rememberof all ChatGPT messages would have to trigger a web search for “one fifth of Google” to hold.OpenAI’s own data puts “seeking information” at 20.2 % of consumer messages. Not impossible, since advice and shopping questions can search the web too, but OpenAI has not shown it.

Sources: OpenAI Signals v3.0, 26 Oct 2026 (CC BY 4.0) · OpenAI usage study, 15 Sep 2025 · Google, 3 Mar 2025 · Big Technology, 3 Nov 2026. Own calculation; rounded.
Powered by Thorsten Meyer AI

Now the arithmetic. This is our own rough calculation, not OpenAI’s and not Google’s:

  1. Google says it handles “more than 5 trillion searches” a year, based on January 2025 data. That is at least 13.7 billion a day. One fifth is at least 2.7 billion a day.
  2. OpenAI’s 2025 usage study counted 18 billion messages a week in July 2025, when ChatGPT had 700 million weekly users. That is about 2.6 billion a day.
  3. Fry now says 1.2 billion weekly users. If each user sends as many messages as in 2025, that is about 4.4 billion messages a day.
  4. 2.7 billion out of 4.4 billion is about 62 percent.

So for the one-fifth claim to hold, roughly six in ten of all ChatGPT messages would have to trigger a web search, three times the share OpenAI itself labels “seeking information.”

Is that impossible? No. ChatGPT also searches the web for advice, shopping, health or finance questions, which Fry mentioned. One question can trigger several lookups. OpenAI may count messages from business accounts that Signals leaves out. And users may now send more messages each than in 2025. But each of those is an assumption, and OpenAI has not published the data that would settle it. Google’s side cuts the other way: five trillion was a minimum in early 2025, and Google says growth has since sped up, which would make one fifth harder to reach, not easier.

What Google says

Google did not answer OpenAI directly. Its own numbers, from Alphabet’s third-quarter results of 28 October 2026, are these:

  • Revenue from “Google Search & other” was $63.35 billion, up 12 percent from $56.57 billion a year earlier.
  • Sundar Pichai said: “our year-over-year query growth rate is the highest it’s been in the last five years.”
  • Alphabet reports query growth only as a rate. It does not publish a daily or monthly total.

That is the strange part of this moment. Both sides say they are growing fast. Both can be right if people simply ask more questions overall, with AI in Google’s own results and in ChatGPT. But neither company publishes the raw counts, so you cannot check either claim to the decimal.

StatCounter shows how little has changed in classic search. In October 2026, Google still had 91.05 percent of the clicks search engines sent to websites in StatCounter’s network, and Bing 4.8 percent. Similarweb’s ranking of the most visited websites for September 2026 still has google.com at number one and chatgpt.com at number five.

Answers without clicks

Here is why the yardstick matters so much. A search engine sends you somewhere. A chatbot answers you where you are.

A wooden signpost with two blank arrow boards at a fork of two gravel paths in a green meadow
Two roads to an answer: a list of links, or a reply in a chat. AI-generated illustration.

Ahrefs puts it bluntly: “Google connects people to websites, ChatGPT keeps them in conversation.” Its October panel shows the gap: for every visit ChatGPT sent to those websites, Google sent about 69.

Google is moving in the same direction with its own AI answers. Pew Research Center tracked the browsing of 900 U.S. adults in March 2025. When Google showed an AI summary, people clicked a regular result in 8 percent of visits. Without the summary, it was 15 percent. Only 1 percent of visits included a click on a link inside the summary.

So even if ChatGPT handles a fifth of Google’s questions, it does not send a fifth of Google’s visitors. For a website, a question answered inside a chat is often a visit that never happens.

And now the ads

The money is the other reason this claim matters. In January 2026 OpenAI announced it would test ads in ChatGPT for users of the free plan and the $8 Go plan in the U.S. Its promise: “Ads do not influence the answers ChatGPT gives you.” Plus, Pro, Business and Enterprise subscriptions do not show ads.

At the New York event, Fry said, as reported by The Verge: “most people do not seem to mind ads in the product,” and “ads have maybe shown us that we could be doing this more, actually.”

A search share of one fifth is a powerful sales line for advertisers. That is a reason to look at it carefully, not a reason to dismiss it. Money pressure on OpenAI is real, as we reported when it reportedly raised $11 billion of a $30 billion goal, and ads are one way it pays for free use, the same tension behind the end of the AI flat rate.

Early independent data on whether those ads work is thin. A working paper by Maximilian Kaiser and Christian Schulze, published in October 2026 and still under review, found that in U.S. traffic a paid ChatGPT visit brought $0.89 in revenue against $2.45 for a paid Google search visit, and converted at 0.37 percent against 1.17 percent. The data came free from an e-commerce analytics firm for which Kaiser is a scientific advisor; the paper says the firm had no role in the research. It is one study, not a verdict.

What this means for companies

Thorsten Meyer AI
What to do

Infographic 03 · Companies and users

Treat AI answers as a second shelf, not a replacement

Search still sends most of the visits. AI answers are where more questions start. Plan for both.

For companies

  • Measure both shelvesTrack visits from Google and from ChatGPT and other assistants every month. Across one large panel, ChatGPT sent 0.36 % of visits in October.
  • Decide on OAI-SearchBotBlocking it keeps your site out of ChatGPT search answers. Blocking GPTBot (training) is a separate setting.
  • Write to be quotedDated facts, one clear answer per page, sources. That helps AI answers and Google alike.
  • Test ads against a controlCompare revenue per visit, not clicks. Early independent data is a single working paper.
  • Ask: share of what?Questions, prompts or clicks? Any “AI search share” without a definition is a sales line.

For users

  • Open one sourceFor health, money, legal or anything you act on, check that the page says what the answer claims.
  • Look for the labelFree and Go users in the U.S. can see ads. OpenAI says ads do not change answers.
  • Use both toolsA chat for a quick overview; a classic search for shops, reviews and official forms.

When Google shows an AI summary, people clicked a regular result in 8 % of visits, against 15 % without one (Pew, March 2025 data).

General information, not professional marketing, legal or investment advice.

Sources: Ahrefs traffic panel, Oct 2026 · OpenAI crawler docs · Pew Research Center, 22 Jul 2025.
Powered by Thorsten Meyer AI

  • Treat AI answers as a second shelf, not a replacement. ChatGPT is a real place where people ask about products, prices and providers. But the click data says Google still delivers most search visits by far. Do not cut search work to chase a chatbot.
  • Decide whether you want to appear in ChatGPT search. OpenAI says a site that blocks its search crawler, OAI-SearchBot, will not be shown in ChatGPT search answers. Blocking its training crawler, GPTBot, is a separate setting. Check what your robots.txt actually allows.
  • Measure the referrals, not the hype. Look at your own analytics for visits from ChatGPT and other assistants each month. If the number is tiny, you are not alone: across Ahrefs’ panel it was 0.36 percent in October.
  • Write for being quoted. Clear facts, dates, prices where you are allowed to state them, and a page that answers one question well are what AI answers can cite. That also helps in Google.
  • Test ads with a control group. If you try ChatGPT ads, compare them against search ads on revenue per visit, not on clicks alone, and keep the test small until you have your own numbers.
  • Ask vendors for definitions. When anyone sells you “AI search share,” ask: share of what? Questions, prompts, or clicks?

What this means for users

  • A search inside a chat is still a search. ChatGPT pulls from the web when your question needs fresh facts. The answer is a summary of other people’s pages.
  • Open at least one source when it matters. Health, money, legal questions and anything you will act on: click the link and check that the page says what the answer claims.
  • Know who pays. On the free and Go plans in the U.S., ChatGPT can show labeled ads. OpenAI says ads do not change the answer. Look for the label anyway, the same as with Google’s sponsored results.
  • Use both. For a quick fact or a list of options, a chat is fast. For comparing shops, reading reviews or finding an official form, a classic search with real links often gets you there quicker.

Honest limits

  • OpenAI’s number cannot be checked. It comes from a spoken remark. OpenAI did not publish the count, the Google baseline, the period or the countries.
  • Our arithmetic is rough. It assumes each ChatGPT user sends as many messages as in July 2025 and uses Google’s 2025 minimum. Change those assumptions and the result moves. We show it to make the question visible, not to replace OpenAI’s number with ours.
  • Two different OpenAI datasets disagree. OpenAI’s 2025 study put “seeking information” at 24 percent of usage by mid-2025; the newer Signals data shows 18.3 percent for June 2025. OpenAI has not explained the gap.
  • Panels are panels. Ahrefs and StatCounter see only websites that use their tools, and StatCounter does not publish the method behind its Google-versus-ChatGPT comparison. Similarweb data is an estimate.
  • Neutrality note. This article was researched and drafted with the help of Anthropic’s Claude. Anthropic competes with OpenAI and Google; Claude appears in Ahrefs’ traffic panel with 0.01 percent. We have kept to published sources and stated where each number comes from.

Not investment advice.

Sources

Researched and drafted with AI assistance, including Anthropic’s Claude.

You May Also Like

Reality Check: If AI Is So Great, Why Isn’t Productivity Skyrocketing?

The truth behind AI’s productivity promise reveals unexpected hurdles that are slowing its widespread impact and raising important questions.

Robot Coworkers: Collaborative Robots Often Need More Help From Humans Than Hyped

Discover why collaborative robots still need significant human assistance despite their hype, revealing challenges that could reshape their role in workplaces.

The New SaaS Competitive Frontier

AIThis post was created with the assistance of artificial intelligence (AI).By Thorsten…

Lovable Was the Most Copied AI Product of 2025. Then a Lobster Changed Everything

AIThis post was created with the assistance of artificial intelligence (AI).A new…