Google’s AI-powered search summaries have become the most widely used artificial intelligence product on earth, reaching 2.5 billion monthly users, according to an investor presentation Alphabet released this week. The figure puts Google’s AI Overviews ahead of every competing AI service by a wide margin and shows how deeply the company has embedded AI into its existing products.
The presentation also disclosed that the standalone Gemini app has crossed 900 million monthly active users, more than doubling in a year. Together the two figures describe a company whose AI reach now exceeds the population of most continents. Gemini, Alphabet said, is embedded in 13 Google products that each serve more than a billion users, including Search, Gmail, Android, Chrome and YouTube.
The numbers point to Google’s most durable advantage: distribution. No rival can match the installed base of a company whose software ships on the world’s dominant phone operating system and whose search engine still handles the majority of global queries. OpenAI’s ChatGPT, by contrast, reaches a few hundred million weekly users, and even the largest competing models must be downloaded or opened deliberately.
But scale is not the same as revenue, and the presentation drew a line between the two. AI Overviews appear at the top of billions of search results each month, yet Google has not disclosed how many of those appearances generate advertising. The company says ads can appear alongside AI summaries, but analysts note that a summary that answers a question directly often removes the need to click a link, potentially shrinking the number of ad impressions.
The monetization question is the central tension of Google’s AI strategy. Traditional search makes money when users click paid results. An AI summary that resolves the query in the page itself can reduce those clicks, even as it keeps the user inside Google’s own ecosystem. The company’s management has argued that overall search usage rises when results are better, and that AI features increase time spent on Google properties, but the unit economics of AI summaries remain unclear.
There is also the question of habit. Users can consume AI Overviews without ever opening a chat interface or paying for a subscription, which makes the 2.5 billion figure impressive and, in a sense, superficial: it measures exposure, not commitment. Analysts who follow the company say the metric most worth watching is not monthly users but whether those users return, and whether the summaries keep people inside Google’s ad-supported world rather than pushing them toward dedicated assistants.
The gap between usage and willingness to pay is where the risks sit. Google has experimented with charging for certain AI capabilities within Workspace and Gemini, but the vast majority of its AI surface is free, supported by advertising. Competitors such as OpenAI and Anthropic sell subscriptions directly, giving them a clearer line between product usage and revenue. Google must prove that mass deployment converts into income growth rather than simply into higher computing costs.
Those costs are real. Every AI summary requires a model inference, and large language models consume more electricity and chips per query than a traditional keyword search. Alphabet’s capital spending has climbed sharply, and the $85 billion financing announced earlier this week for AI infrastructure suggests the company expects the cost curve to keep rising. If ad revenue does not grow at a comparable pace, the most popular AI product in the world could also become one of its most expensive.
Executives have pushed back on that framing, arguing that AI will expand the total amount of commercial activity on Google’s platforms, and that features like Overviews strengthen the company’s grip on the top of the search funnel. They also note that the Gemini app’s 900 million users include people paying for the advanced tier, which gives Google a growing subscription business on top of advertising.
AI Overviews arrived with controversy. When Google first rolled out the feature widely, it produced a string of well-publicized errors, including instructions that ranged from the silly to the dangerous, and the company pulled back while it fixed the underlying model. The corrected version has become a default part of search, and the 2.5 billion figure reflects how quickly the feature was absorbed into the habit of using Google. The episode remains a cautionary example of the gap between reach and reliability.
The scale gap with rivals is widening. ChatGPT, the most used standalone AI assistant, serves hundreds of millions of users a week, a number that is remarkable for a single application but still a fraction of Google’s footprint. The comparison matters for the AI industry’s economics: Google monetizes through advertising at scale, while OpenAI and Anthropic depend on subscriptions, and the two models will produce very different winners depending on which proves sustainable. Google’s argument is that reach wins, and the Overviews number is the evidence it offers.
Analysts who model the economics say the key figure is not users but revenue per query. An AI summary may cost several times more to produce than a traditional search result, and if it also displaces a paid click, the unit math deteriorates. Google has not disclosed enough data to settle the question, and analysts are split between those who believe AI search will eventually lift revenue and those who see a structural drag. The company’s own messaging, emphasizing that Overviews keep users inside Google properties, suggests the internal debate is not fully resolved either.
For now, the scale speaks for itself. Google has won the battle for reach, and no rival appears able to match it in the near term. The open question is whether reach, on its own, pays for the machines that produce it. The answer will determine not only Google’s margins but whether the AI product race ends in a profitable equilibrium or a scramble for the last unsubsidized dollar.
