Children's wellbeing online: social media quantitative and qualitative report

Key Findings

• — Asked to rank the riskiest AI chatbot features, parents in the panel survey placed chatbots offering romantic relationships first (46% top three), followed by chatbots that appear to have real thoughts or feelings (36%) and chatbots offering friendship (34%); 52% of 16-21-year-olds identified a chatbot acting like a boyfriend or girlfriend as a risky feature, with creating harmful content (49%) and confidently wrong answers (47%) close behind
• — 84% of panel parents agreed AI chatbots should have minimum age requirements (54% strongly) and 85% agreed they should have restricted access to certain features (54% strongly); 16-21-year-olds in the panel survey were also strongly supportive (83% and 90%), with the self-selecting public consultation parents higher still (94% and 95%)
• — Among teachers and youth practitioners, 85% agreed AI chatbots should have minimum age requirements (56% strongly) and 86% agreed on feature restrictions (56% strongly)
• — Parents ranked 'having someone to talk to when friends or adults aren't around' last among chatbot benefits for children, cited by only 6%, while 16-21-year-olds rated companionship-related benefits higher (22%)

Methodology Notes

Published on gov.uk 2026-07-15 (content API first_published_at), report PDF 'DSIT Children's Wellbeing Online: Final Report' fetched and read (executive summary, Section 8 on AI chatbots, methodology annex). Nationally representative Savanta online panel surveys: parents and guardians n = 5,010; children and young people aged 10-21 n = 9,017 (AI questions asked to the 16-21 subset); teachers and youth practitioners n = 2,234 (school staff n = 1,218; youth practitioners n = 1,016), weighted post-fieldwork; fieldwork 30 March to 20 May 2026. Two open-link public consultation surveys ran 2 March to 26 May 2026 (parents n = 39,116), which the report states were not subject to quotas or weighting and are not nationally representative. Qualitative strand: deliberative events co-designed with youth collaborators. Attitudinal self-report only; AI is a subsidiary theme within a social-media-focused instrument; commissioned by the department consulting on the policies the survey measures support for. The PDF is watermarked 'OFFICIAL / Classified: Private' despite public publication, a template artefact.

Sources