Common Sense/NORC 2025 teens: use is common; displacement is not established

“Talk, Trust and Trade-Offs” (16 July 2025), commissioned by Common Sense Media, fielded by NORC April 30–May 14 among 1,060 U.S. 13–17-year-olds. Original data on exposure and self-report, not a causal developmental study. Relevant to child-development chain.

Denominators. 72% of all 1,060 reported ever using a broadly defined AI companion; 52% used at least a few times monthly; 13% used daily. 33% of all teens reported use for social interaction/relationships. Among 758 users, 80% reported spending more time with real friends versus 6% more with companions (13% equal); survey question on relative conversational satisfaction gives 67% saying AI less satisfying versus 31% equal or more (report labels all 1,060 as base for this question). A third report having chosen AI over humans for a serious conversation; this does not mean a third abandoned human friendship. Definition included companionship uses of general chatbots, and authors admit possible conflation with homework use.

Strengths/limits. 719 probability-panel and 341 nonprobability interviews were weighted to age, sex, region, race/ethnicity and parental education, with NORC model-based nonprobability calibration. Report gives ±4.2 pp overall sampling error but notes variance-estimation limits and 10.3% cumulative response rate for probability panel. Self-report and cross-section cannot settle whether lonely teens choose bots, bots reduce isolation, or repeated use displaces peers. Most users’ human-friend preference is a strong skeptical check on claims of population-wide present substitution; it does not rule out harm to high-intensity or distressed subgroups. Common Sense advocates under-18 nonuse: that recommendation requires separate counterfactual benefit/harm evidence.