More than a third of Gen Z say they rely on AI for friendship, according to a report from Wired. The number matters less as a curiosity than as a threshold: at that share, talking to a model about your day is no longer an edge case among young adults, it is a mainstream behaviour.
That reframes the companion-AI debate. For most of the last two years the category has been argued about as a risk story — dependency, parasocial attachment, substitution for human contact — usually with the implicit assumption that the users are a small, self-selecting group. A third of a generation is not a small, self-selecting group. It is a cohort for whom the chatbot is simply one of the available social options, sitting alongside group chats, Discord servers and voice channels rather than competing awkwardly with them.
It also changes the commercial picture. Friendship is a high-frequency, low-churn need, and products that meet it daily accumulate enormous usage. The companies building companion models have been treated as a peripheral bet next to coding and enterprise assistants; the usage data suggests the periphery may be where the habit forms first and hardest.
The uncomfortable questions stay the ones researchers have been raising: what happens to social skills that only develop through friction with other people, who is accountable when a companion product behaves badly at the moment a vulnerable user needs a human, and what any of this does to loneliness in aggregate rather than in the moment.
Wired's write-up carries the breakdown of how the figure was arrived at and which behaviours respondents actually described. Read it before treating the one-in-three number as settled.

