ChatGPT at 20% of Google search volume: what OpenAI’s number really counts, and how to get your brand into ChatGPT answers

What did OpenAI actually say about ChatGPT and Google search volume?

The claim came from a fireside conversation, not a filing. On the evening of Monday 2 November 2026, at an OpenAI-hosted dinner in New York, ChatGPT consumer product lead Adam Fry told Alex Kantrowitz of Big Technology that “ChatGPT’s volume of search-like queries is now roughly one-fifth the volume of queries of Google.” Kantrowitz published the exchange on 3 November, alongside three other figures: ChatGPT went “from one billion to 1.2 in just the last two months”, its voice and dictation modes have 250 million combined weekly active users, and users create 1.5 billion images with it a week.

Fry’s own framing of the growth was “roughly 100 million users in 30 days, and another 100 million in another 30 days”, and he called the search figure “a pretty amazing statistic when you think about it, mainly because we started just about a year and a half ago”. 9to5Google reported his definition of a search in one line: “Every time you ask a question and go search the web, that’s a search.” The same report relays, via The Verge, that Fry said “most people do not seem to mind ads” in ChatGPT and that “we could be doing this more, actually”. We covered the ads side in ChatGPT ads in 89 countries.

Three things were not said. OpenAI did not give the number of search-like queries, did not say which Google figure it compared against or for what period, and, as Search Engine Land’s Danny Goodwin noted, did not explain whether several questions inside one ChatGPT conversation count as separate searches. The 20% compares query volume, not share of the search market.

Is ChatGPT really doing 1 trillion searches a year?

Only if you accept two assumptions. Google last disclosed its volume in 2025 as more than 5 trillion searches a year. Search Engine Land’s arithmetic is simple: 20% of that is more than 1 trillion search-like queries a year for ChatGPT. But, as the same piece says, “the companies haven’t established that they’re counting searches the same way”. Apply seoClarity’s 7.5% instead and the figure is around 375 billion.

The gap between the estimates is mostly definitional. OpenAI counts any prompt that triggers a web search, including, in Fry’s words, “finance questions or health questions, where you really want to make sure it’s up-to-date information”. Ahrefs’ February 2026 analysis started from OpenAI’s own figure of 2.5 billion prompts a day (July 2025), which is 18% of Google’s 13.7 billion daily searches, then classified 65% of usage as “things people traditionally searched on Google” to arrive at 11.86%. seoClarity’s October 2026 model scales OpenAI’s published usage categories to 1.2 billion weekly users and assumes 20% growth in Google’s volume, which is where 7.5% comes from.

Is ChatGPT bigger than Bing now?

On OpenAI’s definition, probably. Responding to the claim on X, Kantrowitz added “Fwiw, Bing has 5% search share”, and Glenn Gabe’s reaction, as reported by Search Engine Roundtable, was “So it’s bigger than Bing now? Hmm...”. The caution is warranted. Search share is measured as a fraction of searches across engines, while OpenAI’s 20% is a ratio of ChatGPT’s web-triggering prompts to Google’s queries. Ahrefs’ earlier piece also notes an estimate of 1.2 billion searches a day for Bing, which on raw numbers sits close to seoClarity’s 1.24 billion daily search-like ChatGPT messages. Treat “bigger than Bing” as plausible on OpenAI’s count and unproven on any shared one.

Does ChatGPT send traffic to websites the way Google does?

No, and this is the part the headlines leave out. Ahrefs measured traffic share across 76,000 websites using its web analytics: Google accounted for 39.98% of visits and ChatGPT 0.21%, which is how it reached the finding that Google sends 190 times more traffic than ChatGPT. Dividing traffic share by query volume gave an estimated click-through rate for ChatGPT 96% lower than Google’s. Those are February 2026 figures, before the latest growth, but nothing published this week suggests the ratio has closed. We looked at where the AI referrals that do arrive come from in AI referral traffic vs Google organic.

What Google’s own numbers say the same week

The second half of the story is that Google is not shrinking. Semrush’s State of Search Q3 2026 report, which analysed desktop activity in the US and 32 European countries from July 2025 to September 2026, found the share of US Google desktop searches that led to an organic click on an external website rose from 39.74% in July to 46.11% in September, the highest in the 15-month study. Zero-click searches fell from 24.97% to 22.99%, and clicks to Google-owned properties fell from 17.84% in August to 14.19%. In Europe the organic click share reached 49.56%.

AI is growing from a small base on the same panel. AI tools were 1.29% of US desktop activity in September, up 79% from 0.72% a year earlier, while traditional search was 9.47%, down from 10.29%. Google still took 92.43% of US desktop searches. Google’s AI Mode doubled its share of US desktop activity in the quarter, from 0.15% in July to 0.31% in September. Queries are getting longer: searches of more than 15 words were 2.58% of US Google desktop queries in September, up from 1.67% in August. Rand Fishkin, SparkToro’s co-founder, told the report that this suggests people are treating Google more like an AI assistant. Google itself devoted an Ads Decoded episode on 28 October to the same long-form query trend, with Shivani Mohan, its vice president of data science for Search.

One caution on the Semrush data: it is desktop only, and it measures where sessions start rather than how many questions are asked inside them. It cannot settle the 20% debate. What it does show is that the destination web is still being clicked, and that the clicks increasingly go to specific pages rather than homepages.

What site owners are seeing: fewer homepage clicks and a strange spike

Two field reports this week show how the shift lands in Search Console. In an analysis of 110 homepages across six industries, Previsible’s Jenny Lee Lynch compared Q3 2026 with Q3 2025 and found 81 (74%) lost organic clicks, with a median decline of 21%. SaaS homepages were hit hardest, with 21 of 25 losing clicks and a median drop of 35%. For 63 of the 110 sites (57%), the page earning the most impressions in Google’s AI features was not the homepage. Across 10 of the sites, branded queries drove an average of 79.6% of homepage clicks, and the piece cites Ahrefs data showing AI Overviews on 82.91% of branded keyword searches by 29 September, up from a monthly average of 61.3% in July.

The second report is odder. From 31 October, some sites saw clicks in the Search Console performance report jump several times over while impressions stayed flat. Search Engine Land reported it first, and Search Engine Roundtable gathered the practitioner threads. Google had not commented by 4 November, when most of the affected properties had returned to normal. Two of the most experienced voices in the thread reached the same hypothesis.

Whether or not agents caused this particular spike, it is a preview of a measurement problem: as ChatGPT and other assistants fetch pages on a user’s behalf, some of what lands in your analytics will not be a person. Gabe himself added that most properties did not show the spike and that it “very well could be a bug”. The practical response is in the measurement section below.

What decides whether ChatGPT cites your site?

Start with the mechanics, which OpenAI documents. Its crawler documentation says “OAI-SearchBot is used to surface websites in search results in ChatGPT’s search features. Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links.” It is independent of GPTBot, the training crawler, and of ChatGPT-User, which fetches a page when a user asks for it and “is not used to determine whether content may appear in Search”. OpenAI notes it can take about 24 hours for a robots.txt change to be reflected. We explained the training-side controls, and the CDN settings that silently block these bots, in the AI training control for robots.txt.

Then look at who gets cited. Ahrefs’ 3 November study of 10,000 US brands, using 260 million non-branded prompts across ChatGPT, AI Mode and AI Overviews, found that 46% of brands had zero mentions in ChatGPT, against 23% in AI Mode and 18% in AI Overviews. It then correlated AI mentions with seven kinds of off-site mention. Mentions in Reddit threads correlated most strongly, followed by domain mentions, webpage mentions and news coverage. Ahrefs is careful that these are correlations, not proof of cause, and that the pattern held when brands were grouped by size.

Publishers show the same pattern from the other side. Ahrefs’ companion piece on OpenAI’s publisher partnerships found partner sites’ share of ChatGPT citations among its 200 most-cited domains moved from 7.5% in March 2026 to 15.8% in July and 12.5% in late September, with ten partner domains among the 50 sites ChatGPT cites most. The most-cited partners in its November snapshot of US ChatGPT answers were product and lifestyle titles rather than hard news: vogue.com (42,108 answers), caranddriver.com (41,603), tomsguide.com (36,439), goodhousekeeping.com (35,728) and techradar.com (31,415). Two findings matter for everyone else. The partners ChatGPT cites most are buying-guide and review publications rather than hard news, and its citations churn: in a four-week test of more than 1,300 prompts, only about a quarter of the pages cited for a prompt were cited again the next day, and the most-cited domains changed by about 14% from one day to the next.

How to get your brand into ChatGPT answers: the checklist

1. Confirm OAI-SearchBot can reach you. Check robots.txt for OAI-SearchBot and GPTBot separately, then check your CDN or host’s bot settings, which can block AI crawlers without touching robots.txt. Fetch a key page with the OAI-SearchBot user agent and confirm a 200. Our technical SEO audits include this check.
2. Keep the Google basics, because they are the ChatGPT basics. Google’s own guidance for AI Overviews and AI Mode says there are “no additional requirements” beyond being indexed and eligible for a snippet, and that “you don’t need to create new machine readable files, AI text files, or markup to appear in these features”. Indexable, fast, crawlable pages with structured data that matches the visible text serve every engine.
3. Write answer-first pages for question-shaped queries. Fry’s examples were up-to-date finance and health questions. Build pages that state the answer in the first 60 words, then show the evidence, for the exact questions your customers ask ChatGPT. Our AEO service maps those prompts by category.
4. Earn mentions where ChatGPT looks. The Ahrefs data points at Reddit threads, independent reviews and comparisons, news coverage and YouTube. Participate in the communities where your customers compare options, pitch genuine stories to trade press, and get listed on the review sites your category uses. Buying placements or seeding fake threads is against platform rules and, as Reddit’s COO said on its Q3 call (we covered it here), something the platform is actively working against.
5. Make your brand an unambiguous entity. Same name, same description, same address and founder details on your site, LinkedIn, Google Business Profile, Wikidata where eligible, and the directories of your sector. Ahrefs now tracks brands as entities, and so do the engines.
6. Refresh the pages that win citations. Because ChatGPT’s citations change daily, a dated statistic or a stale comparison loses its place quickly. Put a visible updated date on evergreen pages and revise the numbers quarterly. Our content service runs this on a calendar.
7. Treat homepages as confirmation, not discovery. The 110-site analysis shows AI features surface pricing, product and resource pages more than homepages. Give those pages the internal links, FAQs and proof points the homepage used to carry.
8. Measure before you optimise. Run your 20 to 50 most important prompts in ChatGPT, AI Mode, Gemini and Perplexity each month, log every brand mentioned and every URL cited, and track your share. Pair it with chatgpt.com referrals in GA4 and branded clicks in Search Console.

How to measure ChatGPT’s share of your own demand

OpenAI’s 20% is a global ratio; what matters is your ratio. Three views, read together, give it. First, referrals: GA4 sessions from chatgpt.com, Perplexity, Gemini and Copilot as a share of organic, by landing page and by conversion, remembering that ChatGPT’s click-through rate is a fraction of Google’s so small numbers can still mean large exposure. Second, presence: the monthly prompt panel above, scored as the share of prompts where you are mentioned and the share where you are cited. Third, Google itself: Search Console’s AI features report for impressions, and the branded-query filter for the clicks that homepages are losing. We build this as one dashboard in our reporting service, with a filter for the kind of bot-driven click spike seen this week so it never inflates a monthly number.