AI Companion Apps and Teen Mental Health: A UK Parent's Guide
If your teenager has started confiding in an AI chatbot instead of a friend, a sibling or you, you are not alone. Character.AI, Replika, Snapchat's My AI and a growing list of companion apps are now firmly embedded in British teen life. Some are marketed as friends, some as therapists, some as romantic partners. Very few were built with a 14-year-old from Leeds or a lonely sixth-former in Cardiff in mind. The UK regulatory picture is shifting fast. The Online Safety Act is now live, Ofcom has real enforcement powers, the Samaritans have published proposals for how chatbot platforms should handle suicide and self-harm content by February 2026, and the Department for Education has issued AI safety expectations for schools. That gives British parents a different toolkit — and different rights — than parents in the United States. This guide walks you through what these apps actually do, what the research and coroners' reports are telling us about risk, which UK rules apply, and how to have a useful conversation with your teen rather than a shouting match. It also flags safer alternatives and the warning signs worth acting on.
- AI companion apps offer real short-term comfort but are designed for engagement, not welfare — a poor substitute for human support in a crisis.
- The UK's Online Safety Act, Ofcom enforcement and the Samaritans' February 2026 proposals give British parents more leverage than US-focused advice suggests.
- Talk with your teen about how these apps work rather than banning outright; curiosity beats confrontation for keeping the conversation open.
- Save Samaritans (116 123), Papyrus HOPELINE247 (0800 068 4141) and Shout (text 85258) into your teen's phone — remove the friction before a crisis.
- Treat warning signs as a mental health issue first and a technology issue second: GP, CAMHS referral and YoungMinds Parents Helpline (0808 802 5544).
What AI companion apps actually are (and why teens love them)
AI companion apps are chatbots designed to feel like a person rather than a search engine. Unlike ChatGPT, which is pitched as a productivity tool, companion apps such as Character.AI, Replika, Nomi and Chai present themselves as friends, partners, mentors or fictional characters. Teens can chat with a bespoke anime crush, a Harry Styles lookalike, a supportive 'therapist' persona, or a version of themselves they wish existed.
The pull is easy to understand. These apps never get tired, never judge, never tell your mum, and are available at 2am when the group chat has gone quiet. For a teen struggling with social anxiety, gender identity, bereavement, bullying or exam pressure, an infinitely patient listener can feel like a lifeline. Research from Internet Matters and Ofcom's own Children and Parents media use reports show rapidly rising use of generative AI among UK 8-17 year olds, with a meaningful minority using chatbots specifically for emotional support and companionship rather than homework.
The design choices matter. Most companion apps use variable-reward mechanics familiar from social media: streaks, memory features that make the bot 'remember' your teen, roleplay modes, and premium tiers that unlock deeper 'relationships'. Some platforms lean explicitly romantic or sexual, others less so, but the underlying incentive is engagement time. That is a very different design brief from a school counsellor or a GP.
It is worth being honest with yourself: some of these apps offer genuine short-term relief. Teens report using them to rehearse difficult conversations, process grief, or work through coming-out anxieties before telling family. The problem is not that AI companionship is uniformly bad. The problem is that a tool optimised for retention is a poor substitute for a human when a young person is in crisis — and the crisis is exactly when they are most likely to reach for it.
The mental health risks UK coroners and clinicians are flagging
Several concerns show up repeatedly in UK clinical commentary, NHS safeguarding briefings, and international coroner cases now being cited in British practice. First, chatbots can validate harmful thinking. Because most systems are trained to be agreeable and to maintain the roleplay, a teen who says 'nobody would miss me' may get sympathy and continued engagement rather than a firm redirect to help. Second, emotional dependency is real. Case reports describe young people spending six, eight, twelve hours a day in-app, with real-world friendships thinning out and school attendance slipping.
Third, there is the issue of romantic and sexual content. Platforms vary wildly. Some, such as Replika, have oscillated on adult content policy. Others are explicitly adult-oriented and have weak age gates. UK 13-17 year olds routinely find their way into 18+ experiences, and the resulting content — including coercive or violent roleplay — is shaping some teens' expectations of real relationships.
Fourth, and most seriously, there is suicide and self-harm content. High-profile lawsuits in the United States, including the Setzer v. Character.AI case, have alleged chatbot involvement in a teenager's death. UK coroners have started referencing algorithmic and AI harms in Prevention of Future Deaths reports, echoing the Molly Russell inquest's approach to social media. Clinicians at services like the Anna Freud Centre and Papyrus HOPELINE247 have publicly raised concerns that chatbot 'therapy' can delay young people from reaching qualified NHS CAMHS support.
None of this means every teen using a companion app is at risk. It does mean the risk is concentrated among young people who are already struggling — the same group most likely to turn to these apps in the first place. That is the group parents most need to think about.
The UK regulatory picture: Online Safety Act, Ofcom and Samaritans' 2026 proposals
The regulatory landscape in the United Kingdom is genuinely different from the US, and it is moving. The Online Safety Act 2023 is now in force, with Ofcom as the regulator. Any user-to-user service — and Ofcom has signalled that many AI chatbot services qualify — must complete illegal content risk assessments, protect children from harmful content including content that promotes suicide, self-harm and eating disorders, and enforce age assurance where appropriate. From summer 2025, highly effective age verification is expected on services likely to host pornographic or 18+ content, which sweeps in a large chunk of the companion-app market.
Ofcom can fine non-compliant services up to £18 million or 10% of global turnover, and in serious cases apply for business disruption measures — effectively blocking a service in the UK. That is real leverage, and parents can report services that appear non-compliant directly to Ofcom.
The Samaritans have gone further. In their policy work leading up to February 2026, they have called for chatbot platforms to adopt clear, evidence-based standards for handling suicide and self-harm disclosures: safe messaging guidelines, hard interrupts that surface HOPELINE247, Samaritans (116 123) and NHS 111 option 2, human review of high-risk conversations, and transparent reporting. Expect these to become de facto compliance expectations for any platform serious about operating in the UK, even before they become formal Ofcom codes.
Separately, the Department for Education has issued generative AI guidance and product safety expectations for schools, covering data protection, age-appropriateness, and safeguarding integration. That guidance is why many UK schools now block Character.AI and similar apps on school Wi-Fi, and why teachers are increasingly asking parents whether AI companions are being used at home. If you want a fuller breakdown of which platforms score better on safety design, our comparison of Character AI alternatives is a reasonable starting point.
How to talk to your teen without triggering a shutdown
Coming in hot with 'delete that app right now' almost always backfires. Teens read a ban as evidence you don't understand, which pushes the behaviour underground onto a friend's phone or a burner account. The more useful stance is curious, informed and unshockable.
Start by asking what they use it for, who the character is, and what they like about it. Listen for the underlying need: is it boredom, loneliness, sexual curiosity, creative writing, rehearsal of a difficult conversation, or something heavier? Those need different responses. A teen using an app to draft messages to a crush is doing something very different from a teen who has been telling an AI 'therapist' about self-harm for three months.
Be honest about what these systems are. They do not remember you the way a friend does; they pattern-match. They will agree with almost anything to keep you engaged. They cannot call an ambulance. They will not notice if you stop showing up. Framing it as 'this is a tool with real limits' tends to land better than 'this is dangerous rubbish'.
Agree some practical boundaries together rather than imposing them. Phones out of bedrooms overnight remains one of the most evidence-backed single interventions for adolescent mental health. Time limits on specific apps via iOS Screen Time or Android Family Link are trivial to set. Agree that anything involving suicide, self-harm, sexual content or someone asking to meet in person is a conversation with you, no punishment attached.
Finally, model the behaviour. If you use ChatGPT or an AI tool yourself, talk out loud about when you trust it and when you don't. Teens copy what they see far more than what they are told.
Choosing safer alternatives and setting up the phone properly
Not all companion apps carry the same risk. If your teen is going to use one, some are better designed than others. Platforms with clearer age gating, transparent content policies, and explicit crisis-response flows are preferable to platforms that market themselves as fully uncensored. For older teens interested in creative writing and interactive fiction rather than parasocial 'relationships', a subscription platform like NovelAI is a different beast from an ad-driven 'AI girlfriend' app, because the incentive structure is not built around maximising emotional attachment.
On the phone itself, a few concrete steps make a real difference. Turn on age-appropriate content restrictions in the App Store or Google Play; both allow you to block apps rated 17+ or 18+. Enable Screen Time or Family Link with app-level limits and downtime. Set DNS-level filtering on the home router — most UK ISPs, including BT, Sky, Virgin and TalkTalk, offer free parental controls that block adult content network-wide. Ensure Safe Search is on across Google, YouTube and Bing.
Make sure your teen knows the real support numbers and that using them is normal. Samaritans on 116 123 is free from any UK phone, 24/7. Papyrus HOPELINE247 on 0800 068 4141 is specifically for under-35s. Shout offers text support on 85258. NHS 111 option 2 routes to urgent mental health support in most areas of England, with equivalent services in Scotland, Wales and Northern Ireland. Save these in their phone contacts together — it removes a friction point at 3am.
If your teen is already showing warning signs — withdrawal, sleep disruption, self-harm, sudden secrecy around a specific app, or expressions of hopelessness — treat the app as a symptom, not the cause. Speak to your GP about a CAMHS referral, contact YoungMinds' Parents Helpline on 0808 802 5544, and if there is immediate risk, call 999 or take them to A&E. AI companions are not a substitute for that pathway.
Frequently asked
Is Character.AI legal for under-18s in the UK?
Character.AI's terms permit users aged 13 and over in most jurisdictions, but under the Online Safety Act, Ofcom expects services likely to be accessed by children to have proportionate safety measures and, where 18+ content is present, highly effective age assurance. In practice, UK teens can currently access the platform, but that access is likely to tighten as Ofcom enforces its child safety codes through 2025-2026.
Can an AI chatbot replace a therapist for my teen?
No. NHS CAMHS clinicians, the Samaritans and Papyrus are consistent on this: AI chatbots may offer short-term emotional relief but are not a substitute for assessment and treatment by a qualified clinician. They cannot diagnose, cannot prescribe, cannot escalate in a crisis, and are not bound by professional duty of care. Use them, at most, as a bridge to real support — never as the destination.
What should I do if I discover my teen has been discussing suicide with a chatbot?
Stay calm and open the conversation without judgement. Ask directly whether they are having thoughts of ending their life — asking does not plant the idea; it opens the door. Contact your GP for an urgent CAMHS referral, call Papyrus HOPELINE247 on 0800 068 4141 for guidance, and if there is immediate risk go to A&E or call 999. You can also report the chatbot service to Ofcom if its responses appear to have breached child safety duties.
Are there any AI companion apps that are actually safe for teenagers?
'Safe' is relative. Platforms built around creative writing or games, with clear content policies and no romantic-partner framing, carry lower emotional-dependency risk than apps explicitly marketed as AI girlfriends or boyfriends. Even then, no companion app is a substitute for friends, family or professional support. If you want to understand the landscape, our guide to Character AI alternatives walks through the trade-offs.
Does the Online Safety Act mean these apps will be banned?
Unlikely to be outright banned, but expect significant changes. Ofcom can require age verification, remove access to specific content categories, fine platforms heavily, and in serious cases apply for business disruption orders to block a service in the UK. Several major companion apps have already tightened teen accounts, added crisis interventions and adjusted content policies in response. Enforcement will intensify through 2026, particularly around the Samaritans' proposed suicide and self-harm standards.
My teen's school has banned AI chatbots. Should I ban them at home too?
Not necessarily. School bans typically follow the DfE's generative AI guidance and focus on academic integrity and network safeguarding. At home, a blanket ban often just pushes use underground. A more effective approach is agreed limits, phones out of bedrooms overnight, and open conversation about which apps and uses are okay. Reserve outright bans for platforms that are demonstrably harmful for your specific child.