"X million users" is one of the most commonly cited figures in AI company coverage, and one of the least independently verifiable. Unless a company is publicly traded and subject to audited disclosure requirements, a user count figure is almost always self-reported, with no independent party checking the underlying methodology.
Why self-reported user counts are easy to inflate favorably
Definitions of "user" vary enormously and are entirely within a company's control to choose favorably - anyone who's ever created an account versus someone active in the last 30 days versus someone using the product meaningfully are three very different populations that can all technically be described as "users" depending on which framing suits the current announcement.
Why this rarely gets challenged in coverage
A large, impressive-sounding user number makes for a cleaner story than a caveated one, and challenging the figure requires access and diligence most outlets covering a fast-moving announcement don't have time to pursue before publishing. The number gets repeated, cited, and eventually treated as an established fact through sheer repetition rather than verification.
What a more careful approach looks like
Asking a company directly how it defines "user" for a given figure, and specifically over what time window, is a simple diligence step that a company confident in its methodology will generally answer, and one that a company inflating its numbers will often deflect or answer vaguely.
Why this matters beyond individual company profiles
Aggregate industry narratives about AI adoption speed are frequently built, in part, from a collection of self-reported company user figures using inconsistent, favorable definitions. The industry-wide picture inherits the same verification gap as the individual numbers it's built from.