ChatGPT vs Google Search Volume: What OpenAI’s 20% Claim Really Means

You may have seen the headline this week: ChatGPT now handles about a fifth of Google’s search volume. It sounds like Google is losing ground fast. The short answer is that the chatgpt vs google search volume figure is OpenAI’s own estimate, based on its own way of counting, and nobody outside OpenAI can check it yet.

It also says nothing about traffic to your website. Google is still growing, and it still sends far more visitors to most sites than ChatGPT does. In this post I explain what was said, where the “1 trillion searches” number comes from, why the comparison is shakier than it looks, and what I’d actually track instead.

What OpenAI Said About ChatGPT vs Google Search Volume

The figure came from Adam Fry, who leads consumer product for ChatGPT at OpenAI. He shared it at an OpenAI event in New York on Monday 2 November 2026, in a conversation with Alex Kantrowitz of the Big Technology newsletter. The interview was published on 3 November, and the number spread across tech and SEO news over the next two days.

Here is what Fry said, in plain terms:

• ChatGPT’s volume of search-like queries is now roughly one-fifth of Google’s query volume.
• ChatGPT has grown from 1 billion to about 1.2 billion weekly users in roughly two months.
• Growth is coming from lookups that need fresh information, such as people, places and things, and from finance and health questions where ChatGPT runs a web search to get the answer right.

All of these are OpenAI’s claims. OpenAI has not published the raw numbers, the time period, the countries covered, or which Google figure it compared against. So the 20% is a company statement, not a measured market share.

How OpenAI Counts a “Search”

This is the part most headlines skipped. According to Fry, a search is counted each time a question in ChatGPT triggers a web search. If one chat has ten questions and each one sends ChatGPT to the web, that is ten searches.

That is a reasonable way for OpenAI to measure its own product. But it is not the same unit Google uses. Keep that in mind for everything below.

Where the “1 Trillion Searches a Year” Figure Comes From

Soon after the interview, people started saying ChatGPT handles 1 trillion searches a year. OpenAI did not say that. It is a sum other people worked out.

The maths goes like this. In March 2025, Google said it sees more than 5 trillion searches a year, based on its internal data from January 2025. Take 20% of 5 trillion and you get 1 trillion. Some reports point out that Google itself was at roughly that level around 2012.

The sum is simple, but it stacks guesses on top of guesses:

• The Google number is old. The 5 trillion figure is from early 2025. Google says query growth has sped up since, so the real baseline is likely higher.
• We don’t know which Google number OpenAI used. It could be global, US-only, daily or yearly. OpenAI hasn’t said.
• The units don’t match. As I explain next, one ChatGPT “search” and one Google search are not the same thing.

So I would not repeat “1 trillion searches” as a fact in a client report. At best it is a rough upper estimate built on OpenAI’s own counting. The honest version is: “If you accept OpenAI’s 20% and Google’s 2025 total, you get about 1 trillion a year.”

Why Search-Like Queries Are Not Google Searches

When I explain this to clients, I use a simple example. Someone types “best running shoes for flat feet” into Google. That is one search. They may click a result, or refine the query and search again.

Now they ask ChatGPT the same thing. ChatGPT may run several web searches in the background to build one answer. Then the person asks a follow-up, “which of these are under $100?”, and ChatGPT searches again. Under OpenAI’s definition, that conversation can count several times.

Here are the main reasons the two numbers are hard to compare:

1. One prompt can trigger many searches. AI tools often split one question into several smaller searches. This is called query fan-out. Google counts the person’s query. A fan-out count can grow even if the number of people asking stays flat.
2. Follow-ups count again. A chat is a back-and-forth. Each follow-up that hits the web can be a new “search”, even if it is the same task.
3. “Search-like” is OpenAI’s label. OpenAI decides which prompts count. Outside researchers can’t see where that line is drawn.
4. Different intent mix. Many Google searches are quick navigation, such as typing a brand name to reach a site. Many ChatGPT questions are research tasks that would have taken several Google searches. Neither is “more real”. They are different behaviours.
5. No published data. Without the numbers and the method, nobody can repeat the result.

None of this means the claim is false. ChatGPT is clearly being used for a lot of tasks people used to do in Google. My point is that “20% of Google” describes OpenAI’s counting method as much as it describes user behaviour.

How Other Estimates Compare

Outside companies have tried to size ChatGPT’s search use before. Their methods differ, and none of them is perfect. But they give a useful range. I’ve named the publishers so you can look them up.

The gap between these numbers comes mostly from definitions. When Ahrefs counted only the stricter “seeking information” type of prompt, its figure fell to about 4%. When it counted every prompt, it rose to about 18%. That swing shows how much the answer depends on what you decide a search is.

The studies also use older ChatGPT data than OpenAI’s new claim, and ChatGPT has grown since. So a higher figure today is not surprising. What is missing is a way to check it.

Google Search Is Still Growing

The “ChatGPT is eating Google” story leaves out what Google itself said a week earlier. On Alphabet’s Q3 2026 earnings call on 28 October, Sundar Pichai said: “our year-over-year query growth rate is the highest it’s been in the last five years.” You can read his full Q3 2026 remarks on Google’s blog.

Pichai also said AI Mode queries are on average three times longer than traditional searches, and that one in five AI Mode searches use voice, images or documents. Google did not give total query counts.

So both things can be true at once. ChatGPT can be handling a large and growing number of search-like tasks, and Google can still be growing. People are not simply moving from one to the other. In many cases they use both, for different jobs.

What About Organic Clicks?

There was also a report this week that Google’s organic click share had reached a 15-month high. That cuts against the idea that clicks from Google are collapsing. However, I could not see the underlying data or confirm who collected it and how. So I’m not going to quote any numbers from it.

What I can say is this. Click-share studies are usually based on panels of users, not on Google’s own data. They are useful for spotting direction over time. They don’t tell you what happened on your site. For that, you need your own Search Console data, which I cover below. If you want more on this, see our guide to measuring SEO value in zero-click search.

Search Volume Is Not Website Traffic

This is the part that matters most for site owners. Even if ChatGPT does handle 20% of Google’s volume, it does not send 20% as much traffic to websites.

In the same February 2026 study, Ahrefs looked at its analytics data across about 76,000 websites. Google sent about 40% of their traffic. ChatGPT sent about 0.21%. By that measure, Google sent roughly 190 times more visitors. Those numbers are from early 2026 and will have moved since, but the size of the gap is the point.

There are simple reasons for this. ChatGPT is built to answer the question inside the chat. Links are there for sources and for people who want more, not as the main result. Google still shows a page of links for most queries, even with AI Overviews on top.

In my experience, ChatGPT visitors are often fewer but more ready to act, because the chat has already done the research. That is a pattern I see, not a rule. You should check it against your own data before you plan around it.

What Site Owners Should Actually Track

Headline numbers like “20% of Google” are interesting, but you can’t act on them. What you can act on is your own data. These are the three places I look.

ChatGPT Referrals in GA4

OpenAI says ChatGPT automatically adds utm_source=chatgpt.com to links in its search results. That is stated in OpenAI’s publishers and developers FAQ. It means you can see many ChatGPT visits in GA4 under the session source chatgpt.com.

What I check every month:

• Sessions with source chatgpt.com, and the trend over three to six months.
• The landing pages those visitors arrive on. These tell you which pages ChatGPT cites.
• Key events (the actions you count as success, like a form fill or a sale) from those sessions, compared with organic search.

Remember that some ChatGPT visits arrive with no tag and no referrer, so they show as Direct. Treat your ChatGPT number as a minimum, not a total. Our guide on tracking ChatGPT, Perplexity and Gemini visits in GA4 shows how to set up a channel group and exploration for this. If your numbers look odd, read why ChatGPT traffic can show as Direct.

Search Console Trends

If you’re worried ChatGPT is taking your Google traffic, Search Console is where you check. Look at clicks and impressions over 12 to 16 months, not just the last few weeks.

• Compare the same months year on year, so you don’t mistake seasonal dips for a trend.
• Split branded and non-branded queries. A fall in non-branded clicks with stable impressions often points to AI answers on the results page. A fall in impressions points to a ranking or demand change.
• Look at the question-style queries you rank for. These are the ones most likely to be answered by AI tools.

If Google traffic is stable or growing, the ChatGPT headline does not change your plan. If it is falling, dig into which pages and queries before blaming AI. I usually find a mix of causes, including updates, competitors and changes on the page itself.

Search Console now also has reports for AI features in Google. Our explainer on Search Console’s generative AI reports covers what they show.

AI Visibility

Traffic only shows the people who clicked. Many people read your brand name in a ChatGPT answer and never click. To get a sense of that, you need to look at visibility, not just visits.

A simple way to start:

• Write a list of 20 to 30 questions your customers really ask. Use sales calls, support emails and Search Console queries.
• Ask those questions in ChatGPT, Google AI Mode and any other tool your audience uses. Note whether your brand is mentioned or your site is cited.
• Repeat it once a month with the same list, and keep the results in a spreadsheet.

AI answers change from one run to the next, so don’t panic over a single result. Look for patterns across many questions and months. There are paid tools that do this at scale, but a manual check is enough to start.

Also make sure ChatGPT can reach your content. OpenAI’s FAQ says sites that want to appear in ChatGPT search summaries should not block OAI-SearchBot, its search crawler, in robots.txt.

What I Would Do This Week

If a client or boss sends you the “20% of Google” headline, this is the short routine I’d run before answering:

1. Check your own Google trend. Open Search Console and compare the last three months of clicks with the same months last year.
2. Check your ChatGPT trend. In GA4, filter session source to chatgpt.com and look at sessions and key events over six months.
3. Compare the two. Put Google organic and ChatGPT sessions side by side. For most sites, the gap will be large.
4. List the pages ChatGPT sends visitors to. Make sure they are accurate, current and easy to quote, with a clear answer near the top.
5. Confirm crawler access. Check robots.txt to make sure you are not blocking OAI-SearchBot by accident.
6. Start a monthly AI visibility check. Use a fixed list of customer questions and record where your brand appears.

That gives you a real answer based on your data, instead of a reaction to someone else’s number.

Conclusion

The chatgpt vs google search volume claim is worth knowing about, but it is OpenAI’s own figure, counted OpenAI’s own way. A ChatGPT “search” is each question that triggers a web search, which is not the same as one Google search. The “1 trillion a year” figure is a sum built on that claim and on Google’s 2025 total, not a number OpenAI gave.

Meanwhile, Google says its query growth is the fastest in five years, and it still sends far more traffic to most websites. My advice is to watch both channels in your own GA4 and Search Console data. A good next step is to set up a proper AI channel in GA4 using the tracking guide linked above, so the next headline is easy to check against your own numbers.