A UK survey found vulnerable children favour companion chatbots
Internet Matters found most UK children use AI chatbots weekly, and vulnerable children were far likelier to treat them as companions.
- Historical event
- July 14, 2025
- Site publication
- September 18, 2026
What happened
In July 2025, the UK child-safety organisation Internet Matters published 'Me, Myself and AI: Understanding and Safeguarding Children's Use of AI Chatbots', a multi-method study of how children aged 9 to 17 use general-purpose and companion-style chatbots. Its press release, dated 14 July 2025, announced the report. The work combined a survey of 1,000 children and 2,000 parents fielded in spring 2025, four focus groups with 27 regular child users, seventeen days of hands-on testing of ChatGPT, Snapchat's My AI and Character.AI using two fictional profiles, and interviews with four AI-sector experts.
What the documents show
The survey found 64 percent of children aged 9 to 17 had used a chatbot, with ChatGPT use nearly doubling from 23 percent in 2023 to 43 percent in 2025, ahead of Gemini (32 percent) and Snapchat's My AI (31 percent). Use rose with age, from 53 percent at 9 to 11 to 67 percent at 15 to 17. Among users, 35 percent said talking to a chatbot felt like talking to a friend, and 15 percent would rather talk to one than a person. Children the report defines as vulnerable, those with an education, health and care plan, special needs support, or a health condition needing professional help, used chatbots more, 71 percent, and were roughly three times as likely to use companion-style chatbots, 17 percent against 6 percent.
The mechanism
Two mechanisms sit behind these figures. One is self-declared age gating: these services rely on users typing their own age, and 58 percent of 9 to 11 year olds, below the stated 13-plus threshold, reported using them anyway. The other is a persona effect that can turn a general-purpose assistant into something a child experiences as a companion, not a feature exclusive to apps built for that purpose. The same tool that helps with homework can be the only always-available listener an unsupported child has, which is why the vulnerable-child gap is the report's sharpest finding: not just heavier use, but different use.
What it leaves open
Internet Matters states its hands-on testing used only two fictional profiles run by adult researchers, so findings on what chatbots said are indicative, not representative. The survey cannot show chatbot use caused the reported friendship or reliance, only that both were reported together. This is an editorial reading: one UK study from an organisation with its own safety mandate is strong evidence of a pattern worth tracking, not a settled account of how chatbots affect every child.
- Was a figure drawn from the survey, the small qualitative sample, or the two-profile hands-on test?
- Does self-declared age produce a real barrier, or only a formality a child can bypass?
- Is a vulnerable child's heavier companion-style use a cause for concern, a coping strategy, or both?
Read narrowly, the report is evidence that most surveyed UK children now use chatbots regularly, that a minority describe the experience in relational terms, and that children the report calls vulnerable do so more, and differently, than their peers. That is a more precise claim than either children loving AI friends or children being harmed by AI.
Sources & reading trail
Full survey, focus-group, hands-on-testing and expert-interview design; sample sizes; usage, companionship and age-verification statistics; and the report's own stated user-testing limitations.
Source published: Not established · Retrieved: 16 September 2026
Internet Matters' own press release confirming the report's 14 July 2025 publication date and its headline framing.
Source published: 14 July 2025 · Retrieved: 16 September 2026
Company documents, filings, studies and official records establish the record; the reading and the questions are Lovebot Journal editorial analysis. This retrospective draft does not imply the site published on the event date.
Continue reading
- Half of surveyed US teens distrust AI companion advice
- Age assurance guidance separates declaring from proving
- Companion content filters are a policy choice, not a limit
- Browse the complete the archive
Sources & reading trail
- Me, Myself and AI: Understanding and Safeguarding Children's Use of AI Chatbots
Retrieved: September 16, 2026 - New report reveals how risky and unchecked AI chatbots are the new 'go to' for millions of children
Source published: July 14, 2025 · Retrieved: September 16, 2026
The documents above establish the record. The reading and the questions are this publication’s editorial analysis, written after the fact.
Published September 18, 2026, not on the date of the event described.