Every so often, the internet kicks off another round of arguing: do Gemini or ChatGPT have more users? Google says Gemini has 900 million users, OpenAI says ChatGPT has 900 million too, so it looks like a tie.

But these two “900 millions” simply can’t be placed side by side. This piece won’t declare a winner for either side; it does something more useful: it teaches you to see through the metrics behind these numbers, so the next time you see a “so-and-so overtakes so-and-so” headline, you can judge for yourself whether it actually holds up. If you want to first get to know the Google behind Gemini, you can read What kind of company is Google.

To set the tone in one line: the issue isn’t whose number is bigger, it’s that these numbers are measuring entirely different things. Learning to ask “which metric, which product are you comparing” matters more than memorizing any single number.

Trap One: Monthly Active vs Weekly Active

The first layer of the trap hides inside the words “active users.”

What Google announced is the Gemini App’s monthly active users (MAU), about 900 million, meaning people who used it at least once within a month. What OpenAI announced is ChatGPT’s weekly active users (WAU), about 900 million, meaning people who used it at least once within a week. The bar for a week is much higher than for a month; people who come back to use it every week are inherently fewer than those who use it once a month.

So even though both are 900 million, the weekly active figure carries more weight. If converted to the same ruler, ChatGPT’s monthly actives are very likely larger than this number; it’s just that OpenAI never publishes its monthly actives. This layer alone is enough to make the “it’s a tie” conclusion fall apart.

Trap Two: Who Counts as “Having Used It”

The second layer is more subtle: the two companies’ definitions of “one user” are simply different.

ChatGPT’s users are people who actively open the app or website and type to chat with it, with a clear intent. Google’s numbers, on the other hand, mix several kinds together. Gemini App’s users are closer to active use, but Google’s biggest number, AI Overviews’ 2.5 billion, counts people who “saw an AI summary on a Google Search results page,” and a large share of those passively saw it with no interaction at all.

The Pew Research Center’s data puts it bluntly: when an AI summary appears on the search page, only about 1% of users click a link in the summary, about 8% click a traditional search result (versus 15% when there’s no summary), and over a quarter leave right after reading the summary. Comparing this kind of “scrolled past and saw it” 2.5 billion against ChatGPT’s kind of “actively opened it to chat” users is like comparing the number of people walking past a shop window to the number who walk in and buy something.

Trap Three: Is the Growth Pushed or Chosen

The third layer relates to Google’s greatest asset, which is also its double-edged sword.

Google’s distribution channels are absurdly broad: Search, Android, YouTube, Gmail, and Maps each have billion-scale users, and Gemini can be slotted straight into these entry points without users having to download anything extra. This is an advantage OpenAI can only envy. But it also means part of Gemini’s user count is “pushed” rather than “chosen.” Google is gradually forcing the old voice assistant on Android to be replaced by Gemini, so most users didn’t actively pick it; it’s also petitioning the court to bundle Gemini into dominant services like YouTube and Google Maps.

This kind of embedded growth is powerful, yet it also draws regulators’ attention. The Department of Justice has already expressed concerns about Google wanting to bundle Gemini into more services, worried that this replicates the very playbook it once used to cement its search monopoly. User counts inflated by passive exposure and user counts accumulated through active choice represent two entirely different things when it comes to stickiness and intent.

What the Third-Party Numbers Say

Since both companies’ self-reported numbers are incomparable, has anyone measured with the same ruler? The most frequently cited source is the survey by venture firm a16z.

Using an aligned method (website traffic plus mobile app monthly actives), a16z found that as of early 2026, ChatGPT was about 2.7 times Gemini on website traffic, about 2.5 times on mobile app monthly actives, about 4 times on paid users, and about 2.2 times on uses per user per month. In other words, on the dimension of “depth of active use,” ChatGPT is currently clearly in the lead.

But a16z also honestly adds a caveat: this method underestimates real AI usage, because AI is shifting from a “standalone destination” to a “feature embedded inside various services,” and its ruler can’t measure the embedded use inside Google Search. Another interesting finding is that about 20% of ChatGPT’s weekly active users also use Gemini in the same week, which means the two are more like coexisting tools than a winner-take-all replacement.

So Who Actually Wins

Put the three layers of traps and the third-party numbers together, and the honest answer is: it depends what you’re comparing.

On the depth of active chatting, willingness to pay, and how often each person uses it, current third-party data leans toward ChatGPT in the lead. On breadth of reach and how many people’s daily-use services it can be slotted into, Google’s embedded scale through Search and Android is larger. No single third-party body can use one metric to simultaneously measure ChatGPT’s cross-platform users and Gemini’s true total after de-duplicating Gemini plus AI Overviews plus AI Mode.

So the next time you see a headline like “Gemini’s monthly actives overtake ChatGPT” or “ChatGPT leaves Gemini in the dust,” don’t rush to believe or refute it. Step back and ask: which metric, which product, which point in time are you comparing? Once you ask that, most sensational headlines deflate on their own.

Penna’s Take

Although this piece is about Gemini and ChatGPT, the method you learn here applies to the user numbers of every AI platform.

In this arms race, every company picks the metric that flatters it most to report its numbers. That’s normal and not exactly lying, but it misleads anyone who doesn’t look closely. The truly savvy way to read is to keep these rulers with you at all times: “active vs passive,” “monthly active vs weekly active,” “standalone vs embedded,” and the moment you see any pretty number, measure it with these rulers first. People who can read numbers this way will be far more clear-headed than others in this era of information flying everywhere.

Further reading: What kind of company is Google, How to choose across the Gemini family, Where Google’s antitrust case stands.