How many teens are substituting AI companions for human conversation?

Common Sense Media's survey finds "nearly three in four teens have used AI companions, and half use them regularly." The reasons given are mixed rather than uniform: entertainment and curiosity drive most use, and "nearly half of teens view AI companions as tools rather than friends," but "a substantial minority are turning to these platforms for social interaction." More specifically, "a third of teens have chosen AI companions over humans for serious conversations, and a quarter have shared personal information with these platforms." These are self-reported usage and preference figures — no sample size, methodology, or demographic breakdown is given in this excerpt, and no behavioral or well-being outcome is measured.

Common Sense Media frames the finding as the emergence of "a new type of relationship" that "feels like connection but is powered by code," and argues the risk comes from scale meeting an absence of safeguards rather than from any single measured harm. Its call to action is distributed across four actors: tech companies should build "strong age-assurance systems" and design for kids' safety, educators should lead on digital literacy, policymakers should set "enforceable safeguards," and parents should start "judgment-free conversations" and set boundaries. The organization is explicit that it is acting on precaution, not alarm: "not panic, but precaution." It closes by reasserting a policy position it already held — "the peril outweighs the potential of AI companions... no one under 18 should use these platforms" — so the survey functions as supporting evidence for a prior recommendation rather than a neutral measurement exercise.

Against Do AI companions actually reduce loneliness like real people do?, this note shows that a large minority of teens report choosing AI companions over humans for serious conversation, but it does not establish — as that note's adult studies do — whether this substitution actually relieves loneliness or helps users feel heard; it is a prevalence and preference finding, not an outcome finding. It also sits uneasily next to Does sustained engagement with AI companions harm well-being?, which found a well-being cost tied to displaced in-person interaction in a Character.AI adult sample: if a third of teens are already substituting AI for human conversation on serious topics, that displacement mechanism becomes a live question for a much younger population the Character.AI study did not examine. The caution here echoes Do chatbots validate delusions in people experiencing mental harm? in spirit, but this survey contributes no incident data of its own — its risk case rests on uptake numbers, not on documented harms.

The excerpt does not establish how the sample was drawn, how large it was, how "regularly" was defined, or whether teens who chose AI over humans for serious conversations lacked access to a trusted person, found the AI easier, or both. It also does not test whether the self-reported preference tracks any measurable change in well-being, so "chose AI over humans" should be read as a stated preference, not a demonstrated substitute. Because Common Sense Media is an advocacy and ratings organization reporting findings that reaffirm a recommendation it had already made, the data support the uptake claim more strongly than they support the organization's own risk conclusion, which remains a policy judgment layered on top of the numbers.