What GA4 counts as a user
GA4 keeps two user metrics. Total users covers everyone who triggered any event in the period. Active users covers those who had an engaged session or otherwise met the engagement criteria, and it is the one most standard reports show by default when they say “Users”. Two figures, one label, and the difference is easy to miss.
How GA4 decides that two visits belong to the same person depends on the reporting identity setting. It can stitch visits together using a User-ID you send when someone logs in, Google signals from signed-in Google accounts who allow it, the device identifier stored in a first-party cookie, and modelled behaviour where consent was refused. Because that setting is applied when the report is built rather than when the data was collected, changing it changes historical figures as well as future ones.
Why the user count matters
It is the number clients quote. It appears at the top of dashboards, in board packs and in agency pitches, and it is treated as a count of human beings who visited the website. It is not. It is a count of identifiers that GA4 believes belong to separate visitors, and the same person can appear as several of them.
That gap matters when the figure is used for anything financial. Dividing revenue by users to get a value per visitor, or comparing users against customers in the CRM, produces a number that shifts with a settings change nobody documented.
Where the user count misleads
Someone who researches on a phone and buys on a laptop is two users unless a User-ID ties them together. Someone who clears cookies, switches browser or opens a link inside an app becomes new again. Ad blockers and refused consent remove visitors from the count entirely, so the real audience is always larger than the report.
Users also do not add up. A visitor who appears in two countries or on two devices is counted once in the total and twice in the breakdown, which is why the rows of a report rarely sum to its headline. And when Google signals is switched on, GA4 will hide rows covering small groups to protect identity, so some detail simply will not appear.
How to act on it
Decide one reporting identity for the property, record which one you chose, and leave it alone — a quiet change is the most common reason a trend appears to break. If your site has accounts or a client portal, send a User-ID; it is the only mechanism that genuinely joins one person’s devices.
For trends and for anything you plan to act on, lean on sessions and key events rather than users, because they are counted more simply and survive comparison. Keep users as a directional measure of audience size, state which variant the report is showing, and treat the exact figure as an estimate. Getting those definitions settled at the start is part of a sound GA4 configuration, not something to argue about after the first report.
