ChatGPT and Gemini have each passed a billion users, a mark the two crossed at effectively the same time.
The number carries consequences beyond the competitive scoreboard. An unwanted behaviour appearing in one response out of a thousand stops being a statistic and becomes volume: at a billion people, any small error rate translates into millions of occurrences, some of them in contexts that matter, like health, money or legal decisions.
It also changes the weight of any behavioural change. Adjusting an assistant's tone, restricting a category of answer or swapping the model underneath now affects, at once, an audience larger than any media outlet ever assembled. And that change typically arrives unannounced, because the product is a service, not an installed version.
For anyone building on these platforms, the figure reinforces a familiar risk: the behaviour you contracted today can change tomorrow on someone else's product decision. There is no frozen version of a service, as noted in this edition's coverage of software longevity.
Adoption at this scale also closes the debate over whether these tools are a passing fashion. The remaining question is different: who answers when something goes wrong at that size.