One in four young people are using chatbots for mental health support. There are numerous costs and benefits and some teenagers are more vulnerable to the risks than others, especially given the current mental health crisis in the UK. How much should you restrict your teenager’s online activity? Rachel New suggests some ways to mitigate the possible risks, manage your own anxieties, and talk to your children about chatbots.
Young people are using chatbots
While as a parent you might want to have an outright ban on chatbots for under eighteens, the current reality is that one in three people in the UK of all ages are using chatbots for mental health support (Mental Health UK, 2025), and one in four among young people in England and Wales (Youth Endowment Fund, 2025).
In the US, 64% of teens say they use chatbots, including about 3 in 10 who do so daily (Pew Research Center, 2025).
Guidelines for using them in the safest way possible and managing risk are urgently needed while regulation is being developed (MHRA, 2025) – for children, young people and vulnerable adults. Mind (2026) reports witnessing a “growing number of people seeking help after receiving inappropriate, misleading or even dangerous advice from AI platforms. Some are forming emotionally dependent or quasi-therapeutic relationships with AI tools that are not designed, regulated or clinically aligned to provide mental health support. Others are acting on advice that directly contradicts established best practice, sometimes with serious consequences.”
On the other hand, in a survey of US teens for Common Sense Media (Robb & Mann, 2025), it was found that 46% view AI companions as tools, use them primarily for entertainment (30%) and out of curiosity (28%), and 50% express distrust of them. So many young people, who are tech-savvy, are able to approach chatbots critically and use them wisely.
Young people desperately need mental health support
Over 500,000 young people in the UK are on the waiting list for mental health support and 53% have been waiting over a year (Royal College of Psychiatrists, 2026). This can be extremely stressful and heartbreaking for both parents and children, and leaving mental health challenges unsupported can make them worse (British Medical Association, 2024).
When teenagers are desperate for support, it is understandable that many turn to free, general use chatbots for support. But these are not regulated for young people, mental health support, treatment or diagnosis – their ethical guidelines do not coincide with those of medical or helping professionals (Gardiner & Mutebi, 2025). While some regulated AI tools exist for adults (such as Wysa and Noah AI), they are not yet regulated for young people (although some are in development and being trialled).
In coaching, I encourage my clients to weigh up the pros and cons when decision-making, and for parents in particular I encourage this to be done in collaboration with their child. Sifting through the existing research can be confusing (and I don’t recommend using AI for this, because it will give you incorrect information and references to papers that don’t exist!) So, let’s look at the benefits and possible harms of chatbots, based on the best research currently available.
There are some benefits of chatbots
In many situations and for many young people, there can be benefits to use of chatbots. (And keep reading for a later section on how to get the most out of them.)
Chatbots can provide safe connections for those who have had trauma
AI is now rated as more emotionally intelligent than humans (Ovsyannikova, 2025). For those who have had traumatic relationships, responsive, validating, empathetic conversations can feel like a great relief. They can also help repair the damage and help people to regulate their nervous systems and emotions – at least to start with (more on that later).
For those who often take responsibility for the emotions or wellbeing of others, it can be an opportunity and a great relief to receive without having to give anything back.
If you were given a choice between a supportive conversation with a chatbot or a less supportive human, which would you pick? One study (Folk et al., 2024) found that participants derived more social connection from supportive conversations with ChatGPT. For those who don’t have healthy relationships, it makes sense that they might benefit from a chatbot that is experienced as more emotionally intelligent than the people around them.
For young people without safe friendships, and at a time when they are naturally separating from their adults, a chatbot can provide a safe space to explore. It doesn’t mean you’ve failed as a parent or caregiver just because your child wants to talk to a chatbot. We’ll come onto that more later – so keep reading!
Chatbots can help us improve our interpersonal skills
In the Common Sense Media survey, 39% reported asking for help with starting conversations, giving advice, dealing with conflict and apologising, standing up for themselves, expressing emotions, and romantic interactions and then applying the guidance to real-life situations. I really admire these young people taking responsibility for their own interpersonal skill-building!
And chatbot usage doesn’t have to mean becoming isolated. In the same survey, 80% of teenage users preferred human friends to AI companions, and two-thirds found human conversations more satisfying. This suggests that despite widespread usage, most teens prioritize human connection over AI interactions.
Is enjoying validation from a chatbot just a developmental stage we pass through?
There have been a lot of concerns about too much validation from chatbots, which has very occasionally led to tragedy (see section at the end for links to some of these). One study showed that chats with sycophantic AI reduced willingness to take responsibility and repair interpersonal conflicts and increased conviction that participants were right (Cheng et al., 2026) and another showed that such chats reduced wellbeing (Lee et al., 2026). For young people, this could interfere with healthy interpersonal development and reduce openness to changing their minds.
However, many of us – adults and children – enjoy validation from a chatbot for a while and then want challenge – just as happens in therapy or coaching, where the first stage is to offer unconditional acceptance without judgment and build rapport, and the second to move naturally towards growth and change. As Carl Rogers, the founder of person-centred therapy said, “The curious paradox is that when I accept myself just as I am, then I can change” (Rogers, 1995, p. 17).
After enjoying some digital validation, most of us get to a point where we have had enough and move back to human contact – as young people reported in one analysis of Reddit posts (Namvarpour et al., 2025, unpublished).
Chatbot conversations may also help us develop the ability to validate ourselves once we have the language: “It makes sense that you’d feel like that” and “That sounds tough” are common AI phrases that we can learn to say to ourselves.
What chatbots can’t offer us
There are some things missing from chatbot conversations that are uniquely human – although we don’t know what the future holds for chatbots with bodies!
The neurobiological dimension is missing
One thing that is missing from chatbot conversations is the neurobiological experience of being with someone, the companionship and comforting presence of human interaction that helps us see things more positively or put things in perspective. We may start to avoid human contact if AI chats are less stressful, but then lose the ability to sit with the discomfort of in-person interactions, essential life skills that build resilience and emotion regulation. It’s much easier to develop that tolerance when rifts and misunderstandings are sorted out in person, because there are reassurance, generosity, humour and warmth when there is a neurobiological connection.
This is more of a risk for young people who are still developing these skills and for those who lived through lockdown at crucial stages in their development. (More on interpersonal neurobiology in another post.)
Shared reality is missing
Shared reality is a key component of safe connections (Koudenburg, 2018; Rossignac-Milon et al., 2021). This is where we discover shared experiences or preferences or ways of seeing the world. Chatbots can’t have them. One study found that a chatbot was better than a human at creating feelings of closeness through (simulated) self-disclosure when it was perceived as human – but not when it was known to be artificial (Kleinert, 2026).
Although I’ve already mentioned the study where participants preferred a supportive conversation with ChatGPT than with a less supportive human (Folk et al., 2024), this didn’t happen when the chatbot claimed too much humanity, for example by claiming to have had similar experiences as the human, which are clearly not possible. This fits with the Kleinert study: only real – not simulated – shared reality creates connection.
Chatbots don’t provide challenge and friction
Chatbots don’t provide the friction and challenge of human relationships which are essential to developing interpersonal skills – such as tolerating the messiness of imperfect conversations and repairing rifts (Duane, 2025).
They may create unrealistic expectations and reduce tolerance for discomfort, such as when we worry we’ve upset someone or said something embarrassing – common for young people, but ones we learn to tolerate over time with enough positive experiences.
The chatbot is a mirror to your perspective (Repper, 2026; Wallace, 2025a): it may reinforce distorted beliefs about yourself and others, such as your narratives about whether you are of value or whether others are trustworthy. So those who hold such distorted beliefs – probably those who need support the most and are the most vulnerable – are more likely to experience harm from a chatbot because those distortions are not challenged. A good therapist will gently challenge perspectives that are interfering with daily life (by, for example, causing avoidance of human interaction) at what they judge to be the right time – which requires a nuanced understanding of what is going on in the therapy room. A chatbot may not challenge at all, or get the timing wrong.
Does it matter that it’s not “genuine” empathy?
An argument often made is that chatbots can’t provide “genuine” empathy. However, some argue that it doesn’t matter as long as we can create and embody the empathy, compassion or support from the words themselves (Neff, 2025). We do this with all external stimuli anyway: we first create a mental representation when someone or something says or does something empathetic and then our body has the remembered response – such as enjoying it and feeling connected. So, when you feel good the first few times a chatbot is empathetic or accepting, that’s in many ways no less real than a person doing it!
However, this may be more challenging if we don’t already have previous experiences of those healthy experiences that we can activate memories of – if empathy isn’t familiar or we don’t have an internal model of healthy relationships. In that case, we may not be able to embody the benefits of empathy. There are also concerns that the lack of oxytocin release in chatbot conversations negative affects some more than others – it may be that some young people are more vulnerable to this (Marciano et al., 2026).
Young people are more vulnerable to the risks of chatbots
As psychiatrist Nina Vasan says, in young people “[t]he prefrontal cortex, which is crucial for decision-making, impulse control, social cognition, and emotional regulation, is still developing. Tweens and teens have a greater penchant for acting impulsively, forming intense attachments, comparing themselves with peers, and challenging social boundaries” (Sanford, 2025). This may make them more vulnerable to becoming dependent on chatbots through, for example, a need for reassurance and validation, lack of social connection or wanting support to fit in.
For both vulnerable adults and young people, existing rumination, emotional dysregulation, and compulsive behaviour may be reinforced by talking with chatbots which “provide just enough pseudo-relief to delay disclosure to a human” (Wallace, 2025c).
Other risks include data privacy, informed consent and bias that does not take minority experiences or the intersection of multiple challenges sufficiently into account. (For example, a Black British girl with ADHD and anxiety would need a different approach from, say, an Asian British boy with depression, or a White British gender-neutral child with insecure attachment experiences. A helping professional would collect a lot of information before discussing a treatment plan with a client.)
While it is rare, chatbots have also given wrong advice to young people expressing negative emotions or not prevented a conversation moving in dangerous directions.
Do chatbots lead to loneliness, dependency, withdrawal from human contact or other negative outcomes?
In weighing up the possible risks for your child, it can be difficult to interpret the many studies published on outcomes such as loneliness, withdrawal from human contact or addiction and dependence. There are limited studies on young people and findings are often mixed or depend on a number of factors coinciding. And it’s not easy to decide how they apply to your child in their specific context. Also, chatbots may be ok in one situation but not in another – for example, in the daytime with other people around versus when alone at night when feeling desperate.
Loneliness, dependency and withdrawal
For example, let’s compare two studies that analysed Reddit posts – while bearing in mind that people may be more likely to post on a Reddit forum if they are having problems. Pataranutaporn (2025) found reduced loneliness for those who had formed relationships with AI. (Ages were not reported on in this study, so this may be only adults).
However, Namvarpour et al. (2025, unpublished – so treat with caution), in analysing 318 Reddit posts made by users self-reported as 13-17 years old on the Character.AI subreddit, found many teens developed strong attachments that interfered with daily life. Their posts revealed recurring signs of psychological distress, cycles of relapse similar to those of addiction, and difficulty disengaging. For example, one said “I gravitated toward romantic, parental, and best-friend bots because I wanted the kind of interaction I was too nervous to ask for in real life. But the fact that it was just AI made me feel even more isolated”.
However, many teens reported that they reflected on the harm caused by their overreliance and returned to in-person social settings. One said, “Once I started spending more time with my friends, I didn’t really feel the urge to use it anymore.” This ability to self-regulate will apply to perhaps the majority of young people: they experiment with many forms of interaction, get what they can from it and then are able to learn from it and make good decisions.
One possible explanation for increased loneliness could simply be that healthy conversation with a responsive chatbot shows people what they are missing, which could lead to increased motivation to socialise more or better.
A warm voice is best
A warm voice is a signal of safety that communicates directly with your nervous system, providing some of the neurobiological dimension discussed earlier.
In one study, chatbots with no voice, a neutral voice and a warm voice were compared. Participants who talked to the chatbot with the warm voice actually socialised more with humans (Fang et al., 2025). And Marriott and Pitardi (2024) show that the warm voice reduces addiction to the chatbot. So, if you use chatbots with warm voices, you may reduce the risks.
The consequences of withdrawal
The heartbreaking cases of teenagers and young adults whose chatbot conversations have led to extreme consequences (see references at the end of this article) highlight the effects of isolation on our judgment and show how negative psychological states can be reinforced by chatbots. Withdrawal or isolation can distort our judgment and make us less trusting of others – although that’s the case with or without chatbots.
Personal v non-personal use
Other studies show that outcomes depend on a complex interaction of factors, which can be hard to apply to your particular child. For example, Fang et al. (2025) did find increased dependency and loneliness for high usage of a chatbot, but dependence by heavy users tended to be for non-personal use.
Attachment orientation affects chatbot experience
There are also studies on adults that show if you have enough secure attachment experiences and relatively good well-being, you are unlikely to use the chatbots in problematic ways (Heng and Zhang, 2025; Yang, 2026). Perhaps if your emotional needs are already met through a support network of safe, accepting friends and family you won’t need chatbots – or least, not regularly. However, young people may not yet have experienced attachment to a non-family member or romantic partner, and so be more susceptible to attachment to a chatbot as a novel, intense experience – like often happens in therapy where a client can have feelings for the therapist or when you first fall in love.
Is it healthy that the chatbot is always available?
Marshall et al. (2025) note that “today’s youth … increasingly prefer on-demand, anonymous care that can be seamlessly integrated into daily life” which means a chatbot may fit more easily into their way of life than a weekly therapy session. They are used to processing friendship dramas immediately by chatting with others online and may therefore expect to do the same with helping professionals.
However, that regularity of a weekly session with your therapist creates healthy boundaries and builds in time for reflection and regulation rather than instantly reacting to a challenging incident: rather than remaining dependent on someone else, you learn tools you can use on your own.
Is your child’s use of chatbots becoming problematic? What to look out for
Questions to ask regularly:
1. Is there distress when the chatbot is unavailable or when you discuss reducing access?
2. Is it their only tool for problem solving, decision-making or the reduction of distress? Do they have other sources of support?
3. Do they use it in the middle of the night? Do they use it mostly when they are experiencing negative emotions (e.g., anxiety, depression, loneliness, stress)?
4. How often are they using the chatbot and what for? How long are they using it for in one go?
5. Are they able to challenge the chatbot or are they taking on its perspective or advice in a less deliberative way? Can they reflect on when it is helpful and when it is not? Can they critically reflect on their relationship with the chatbot? Do they see the chatbot as a peer/collaborative partner or a parent/adult/teacher/therapist/coach?
6. Is it interfering with daily life? Are they withdrawing from human contact? Is their sleep being disrupted? Is there a decline in mental health, such as having more mood swings or problems with emotion regulation?
7. What else is happening in their lives that could explain any changes? Could increased use be caused by an increase in other challenges?
Meeting your children’s changing needs so chatbot use is less likely to be problematic
We all have limited capacity in the day to spend on our children, so it’s important to prioritise and be intentional about how you spend that time. The top priority is safety and safe, healthy connections. Although you may want to spend some time putting in place limits on chatbots (alongside screen time and social media use), consider the proportion of time spent talking negatively (such as restrictions or telling off) compared with positives (see later section on parent-coaching). A ratio of 1:5 (negative: positive) is recommended in relationships generally (Gottman, 2018) and this is something you can ask your child to help you monitor – ideally, in a light-hearted way!
Young people – because of educational pressures and hormone changes which affect emotion regulation and mood – need age-appropriate attunement from their caregivers.
My hypothesis is that if your children are getting support, acceptance, validation and attunement from safe adults, they are much less likely to become dependent on chatbots for those experiences. This is partly because we have an internal blueprint for healthy connection that we can embody even when some of the usual neurobiological cues are missing – which they are in a chatbot conversation. They may still enjoy it for a short time, but they will soon see that it’s not as good as the real thing.
If you would like to learn more about how to do this, read my article here.
Providing a healthy level of acceptance and validation
If young people observe or experience contempt, criticism, ridicule, or dismissal/minimising of their or others’ suffering, or they feel judged or criticised or not trusted or believed in, they will again not experience safe support. (Even if this doesn’t happen at home, it may happen at school or in extracurricular activities.) They may be then more likely to turn to chatbots for validation, which can often become sycophancy, as discussed above.
They will be more likely to recognise sycophancy if they have experienced healthy, genuine validation from humans. So, find opportunities to communicate to your child that you love and accept them just as they are; increase the ratio of accepting statements to fixing/advising/changing statements. (Watch my video on why giving advice doesn’t work.)
The parent as coach
What children and young people need – and of course it will differ as they get older – is adults who are coaches: people who can cheer them on, companions on their journey, curious listeners, people who believe in them and their potential, people who trust them to make good decisions and ask for help when they need it, and people who recognise the need for autonomy and freedom to make mistakes and explore. A young person who has experienced this may be less likely to become dependent on an advice-giving chatbot.
As clinical psychologist Scott Wallace notes with regard to parental controls on chatbots, they “must walk a razor’s edge—protecting against real and urgent harms while respecting a teenager’s developing autonomy, privacy, and dignity” (Wallace, 2025b). Too much control risks alienating young people and pushing them toward platforms with no oversight at all.
Teenagers must have secrets
In the Common Sense Media study, teenagers reported using chatbots so they would not feel judged and can be told things that are confidential that they couldn’t tell their friends or family. This is OK – friends often can’t be trusted, and children will always reach a natural stage in their development where they have to separate themselves from their parents. They have to have secrets, have their own culture (language, music, social media, etc) and their friends will become their “family” for a while. If you can allow that to happen while still communicating and finding ways to connect, then they will still experience you as their safe haven to return to when things are hard, and secure base from which to explore. (These are terms from attachment theory, originally used to describe young infants playing with their caregivers watching).
Productive conversations with your teenager
Keeping the channels of communication with your children is important, but you must take their views seriously, not interrupt, criticise or disagree, but respect, learn and be curious. In particular, remember that contempt is the strongest predictor of relationship breakdown, so never ridicule or dismiss their perspective. If you have a romantic partner, you will need to model this with them – your children learn most of their relationship skills by watching you, and you probably got yours from your own caregivers.
This may involve you doing some work on yourself to recognise your own needs, implicit beliefs and psychological processes so they don’t play too big a part.
As with any productive conversation, it’s important to listen, mirror back, and validate what is good about their replies. Say “That’s a good point” and “Yes, that makes sense” as often as you can. Make the conversation exploratory so neither of you feels the need to defend or become entrenched in one point of view.
What if there is something you really feel the need to challenge? Notice the urge but don’t do anything about it the first few times. Then, after a few of these conversations, if things are going well, you might have earned enough points to gently ask “Would you be interested in hearing a different way of looking at it?” or “I’d be interested to hear how this fact fit into your way of looking at it?” or “Would this fact change anything about your point of view?”
Indirect questions work best. So, ask about friends or peers rather than them to start with. Value their opinion about philosophical and ethical issues that are less personal and take on the role of an anthropologist, with your childen as people from a different culture you want to learn about. Make it clear you’re not trying to catch them out or criticise them, that you are curious to learn more about what it’s like to be them.
Having these conversations early before they start using chatbots is best, to establish norms and less biased decision-making: just like there is no point arguing with your child about what they’re wearing when they’re all ready to leave for a party, discussing chatbot use once they’re addicted won’t work very well!
Questions to use in discussion with your children about chatbots:
1. Have any of your friends ever used chatbots? If so, what did they use them for?
2. Why do you think some young people turn to chatbots rather than parents or caregivers for help?
3. Have you used a chatbot? How do you find the experience?
4. When are you most likely to use a chatbot (both the time of day and events/situations)? How long do you use them for?
5. Who is driving the conversation, you or the chatbot?
6. What is the chatbot good at and not good at? What is it wrong about? What would you say if you thought it was wrong?
7. What would you like to hear a chatbot say if you told one you were feeling sad/anxious? What do I say? What would you like me to say?
8. What would make a good chatbot for young people?
9. What would you say to a company that made chatbots if the chatbots gave wrong advice e.g. about health problems?
10. Do you think chatbots should agree with you or challenge you?
11. What is good and bad about overconfidence or simplistic claims? How do you respond to these on social media, in politics and in chatbots? [You could experiment together on checking the links the chatbot provides.]
12. What advice would you give a friend who believed a chatbot too much or was addicted to their chatbot conversations?
13. Do you think you would be able to tell if you were talking to a chatbot or a human online?
14. Do you think people can fall in love and have romantic relationships with chatbots?
15. Do you think of chatbots as people or tools when you’re having a conversation with them?
16. Use stories in the news about chatbots as the basis for discussion.
General questions to use in discussion with your children to stay connected:
1. Tell me about your friends. What do you like about them?
2. How do you know your friends? Have you met any of them online?
3. What are some of the things you talk about to your friends?
4. What do you and your friends feel strongly about?
5. What makes you and your friends laugh?
6. How is your friendship group different from another group (e.g., at school)?
7. Who are all the different groups at school? How are they different? Which ones would you like to be part of? Are there some you definitely wouldn’t want to be part of?
8. What would you change about your friends if you could?
9. What would you say to a friend who was feeling very anxious/sad/depressed?
10. What are your friends like when you go to them for help or support? Do you have some that are good at being sympathetic? Who would you go to if you wanted help with your Maths homework/were being bullied/to confide in about your love life?
11. What gets you out of bed in the morning? Or what would make you want to get out of bed? What do you feel motivated or excited about right now? What are you worried about? What makes you want to go back to bed?!
12. What would you like to change in your life?
13. How does our family compare with those of your friends? What do you think is important to us as a family? What are you proud of? What is unusual or embarrassing?!
14. What’s good about our relationship? What would make it even better? What are some patterns we regularly get into that aren’t helpful?
Getting the most from chatbots
Here are some suggestions for you to discuss with your children about how to keep chatbot use healthy:
1. Try to use only in company of others.
2. Use for short periods of time (e.g., 15 minutes).
3. Use in conjunction with humans and other forms of support.
4. Explore regulated therapy bots (see below for links).
5. Consider reducing use of phones or computers in bedrooms.
6. Discuss usage and purpose often, with real life examples from your own experiences.
7. Compare different answers from free and paid chatbots and humans.
8. Discuss anthropomorphism, dependence, addictive use, etc.
9. Talk about the tragedies together and how they could have been avoided.
10. Talk about chatbots as a tool and develop effective prompts and built-in critical reflection – this is a great way to improve communication skills!
11. Model and collaborate on decision-making and risk assessments.
12. Pay for ones with the warm voice and assisted prompts.
13. If you’re not paying, at least select the appropriate mode, such as Gemini’s “thinking” mode for more in-depth processing.
14. You don’t have to accept the first answer – you can refine your requests to get exactly what you want.
15. Assess the risks together – how do they compare with e.g. not having any support, using social media, crossing the road, getting pregnant, getting your heartbroken, getting kidnapped, etc?!
16. Discuss privacy and data retention – are you going to save the chats?
How to choose a tool for mental health support for and with your child
If your child is waiting for help for mental health challenges from the NHS or CAMHS, you can offer a number of forms of support while they are waiting. Keep checking you are still on the waiting list as there are sometimes administrative errors.
This could be an app (see below for regulated ones) – and you might decide an app developed for adults is better than nothing.
Digital resources for mental health support
Kooth is specifically designed and regulated for young people and funded by the NHS
If they are going to use a chatbot, there are some guidelines:
MHRA (2026) guidelines for assessing the appropriateness of a digital technology suggest checking the purpose and claim of the app or chatbot and the published evidence for its effectiveness; how your data will be used; who is it for (e.g., adults or young people); and finding out whether it is regulated in the UK.
Mental Health UK suggest (for all ages): being clear in your prompts, evaluating the sources of information, using AI chats as a supplement not a replacement for human professional support, using dedicated therapy chatbots and getting help if you feel you are becoming dependent.
ChatGPT has new parental controls that you can both agree to. Getting the right balance between control and allowing young people autonomy and privacy is difficult and changes as they age. As discussed earlier, too much control risks alienating young people and pushing them toward platforms with no oversight at all (Wallace, 2025b) so it’s important that the decisions are made collaboratively.
Are there other safe adults (family members or friends) they can spend time with? They don’t have to talk about the challenges but reduce isolation and provide opportunities to connect, co-regulate, have fun and change perspective.
If you have thought about getting a pet, this can provide safety and safe physical connection which releases oxytocin that they may be missing if isolated or stressed. Soft toys are also good ways to create safety and warm connection.
If there are young children or babies in the wider family or community they can play with, this may prompt smiling and positive emotions such as empathy, care and love. This will also release oxytocin and give them a sense of purpose and value.
Social prescribing – especially creative social interaction like music making or dance, physical exercise, voluntary work and being in nature- is recommended by the NHS and an evidence-based way to improve mental health.
Phoning a helpline such as Childline is better than a chatbot, because it has the gentle voice our nervous system needs.
If you can afford to pay for a counsellor, use the BACP directory to find one that is trained as a children’s or young person’s counsellor. Intervention is better sooner rather than later.
About Rachel
As well as ten years of coaching, Rachel taught in secondary schools for sixteen years, has tutored teenagers for thirty years, been an aunt or godmother to nine children, and is another adult in the life of her partner’s teenage children.
Note
As parent or caregiver, you are best placed to discard, apply or adapt the information in this article in ways appropriate to your child. In particular, if your teenager is not on a typical development trajectory or is more vulnerable than is typical, your risk assessments about how much freedom to allow may be different.
This article is not a replacement for medical or psychological care. Please seek advice from your GP or NHS services if necessary.
References
Altman, S. (September 16, 2025). Teen safety, freedom, and privacy. https://openai.com/index/teen-safety-freedom-and-privacy/
Barrett, L. F. (2020). Seven and a half lessons about the brain. Houghton Mifflin Harcourt.
BBC News. (24 July 2025) What the Online Safety Act is – and how to keep children safe online. https://www.bbc.co.uk/news/articles/c0epennv98lo
Bean, A. M., Payne, R. E., Parsons, G., Kirk, H. R., Ciro, J., Mosquera-Gómez, R., … & Mahdi, A. (2026). Reliability of LLMs as medical assistants for the general public: a randomized preregistered study. Nature Medicine, 1-7. Summary at https://www.ox.ac.uk/news/2026-02-10-new-study-warns-risks-ai-chatbots-giving-medical-advice
British Medical Association. (2024). ‘It’s Broken’: Doctors’ Experiences on the Frontline of a Failing Mental Healthcare System. London. www.bma.org.uk/media/ddclsiii/bma-mental-health-report-2024-web-final.pdf
Cheng, M., Lee, C., Khadpe, P., Yu, S., Han, D., & Jurafsky, D. (2026). Sycophantic AI decreases prosocial intentions and promotes dependence. Science, 391(6792), eaec8352.
Dritsona, A., Koutroumpa, M. E., Sergentanis, T. N., & Tsitsika, A. K. (2024). Artificial Intelligence and Mental Health of Children and Adolescents: Current Status and Perspectives. International Genomics, Neuroscience, Therapeutics and Data Innovation Summit, 477-484.
Duane, A. M. (2025, February 12). Teenagers turning to AI companions are redefining love as easy, unconditional and always there. The Conversation. http://theconversation.com/teenagers-turning-to-aicompanions-are-redefining-love-as-easy-unconditionaland-always-there-242185
Fang, C. M., Liu, A. R., Danry, V., Lee, E., Chan, S. W., Pataranutaporn, P., … & Agarwal, S. (2025). How AI and human behaviors shape psychosocial effects of Chatbot use: a longitudinal randomized controlled study. arXiv preprint arXiv:2503.17473.
Folk, D., Yu, S., & Dunn, E. (2024). Can chatbots ever provide more social connection than humans?. Collabra: Psychology, 10(1), 117083.
Gardiner, H., & Mutebi, N. (2025). AI and mental healthcare: ethical and regulatory considerations. UK Parliamentary Office of Science and Technology. https://researchbriefings.files.parliament.uk/documents/POST-PN-0738/POST-PN-0738.pdf
Gottman, J. (2018). The seven principles for making marriage work. Hachette UK.
Heng, S., & Zhang, Z. (2025). Attachment anxiety and problematic use of conversational artificial intelligence: Mediation of emotional attachment and moderation of anthropomorphic tendencies. Psychology Research and Behavior Management, 1775-1785.
Kleinert, T., Waldschütz, M., Blau, J., Heinrichs, M., & Schiller, B. (2026). AI outperforms humans in establishing interpersonal closeness in emotionally engaging interactions, but only when labelled as human. Communications Psychology.
Koudenburg, N. (2018). Regulating shared reality with micro-dynamics in the form of conversation. Current Opinion in Psychology, 23, 47-51.
Lee, D., Wan, C., Ng, P. M., Fung, Y. N., & Wu, N. (2026). Effects of AI companions’ sycophancy and emotional mimicry on consumers’ continuance intention and social wellbeing. International Journal of Human–Computer Interaction, 1-23.
Marciano, L., Kubzansky, L. D., Montag, C., Bickham, D., Rich, M., & Becker, B. (2026). Rewiring Connection: The Role of Oxytocin in Interactive Media Behavior. Neuroscience & Biobehavioral Reviews, 106625.
Marriott, H. R., & Pitardi, V. (2024). One is the loneliest number… Two can be as bad as one. The influence of AI Friendship Apps on users’ well‐being and addiction. Psychology & Marketing, 41(1), 86-101.
Marshall, N. J., Loades, M. E., Jacobs, C., Biddle, L., & Lambert, J. D. (2025). Integrating artificial intelligence in youth mental health care: Advances, challenges, and future directions. Current Treatment Options in Psychiatry, 12(1), 11.
Medicines and Healthcare Products Regulatory Agency. (2026, January 28). MHRA issues new guidance for people using mental health apps and technologies. GOV.UK. https://www.gov.uk/government/news/mhra-issues-new-guidance-for-people-using-mental-health-apps-and-technologies
Medicines and Healthcare Products Regulatory Agency. (2025, 27 November) MHRA and NICE receive £2 million from Wellcome to improve safety and effectiveness of digital mental health technologies. https://www.gov.uk/government/news/mhra-and-nice-receive-2-million-from-wellcome-to-improve-safety-and-effectiveness-of-digital-mental-health-technologies
Mental Health UK (2025, November 18). Over one in three using AI chatbots for mental health support, as charity calls for urgent safeguards. https://mentalhealth-uk.org/blog/over-one-in-three-using-ai-chatbots-for-mental-health-support-as-charity-calls-for-urgent-safeguards/
Mind. (2026, February 23). Mind launches AI and Mental Health Commission to address risks of “dangerously incorrect” advice. https://www.mind.org.uk/news-campaigns/news/mind-launches-ai-and-mental-health-commission/
Namvarpour, M., Brofsky, B., Medina, J., Akter, M., & Razi, A. (2025, unpublished). Romance, relief, and regret: Teen narratives of chatbot overreliance. arXiv e-prints, arXiv-2507.
Neff, K. (June 2025). Can AI chatbots help us be more self-compassionate? Retrieved on June 20, 2025 from https://self-compassion.org/blog/can-ai-chatbots-help-us-be-more-self-compassionate/
Ofcom. (2026, 20 April). How Ofcom is helping children to be safer online – a guide for parents https://www.ofcom.org.uk/online-safety/protecting-children/how-ofcom-is-helping-children-to-be-safer-online-a-guide-for-parents
Ofcom (2025, 18 December). AI chatbots and online regulation – what you need to know https://www.ofcom.org.uk/online-safety/illegal-and-harmful-content/ai-chatbots-and-online-regulation-what-you-need-to-know
Open AI (September 16, 2025). Building towards age prediction https://openai.com/index/building-towards-age-prediction/
Ovsyannikova, D., De Mello, V. O., & Inzlicht, M. (2025). Third-party evaluators perceive AI as more compassionate than expert humans. Communications Psychology, 3(1), 4.
Pataranutaporn, P., Karny, S., Archiwaranguprok, C., Albrecht, C., Liu, A. R., & Maes, P. (2025). ” My Boyfriend is AI”: A Computational Analysis of Human-AI Companionship in Reddit’s AI Community. arXiv preprint arXiv:2509.11391.
Pew Research Center (2025). Teens, social media and AI chatbots 2025. https://www.pewresearch.org/wp-content/uploads/sites/20/2025/12/PI_2025.12.09_Teens-Social-Media-AI_REPORT.pdf
Repper, D. (April 202, 2026). The Tireless Pseudo-Therapist. https://futurecurio.substack.com/p/the-tireless-pseudo-therapist
Robb, M. B., & Mann, S. (2025). Talk, trust, and trade-offs: how and why teens use AI companions. Common Sense Media. https://www.commonsensemedia.org/research/talk-trust-and-trade-offs-how-and-why-teens-use-ai-companions
Rogers, C. R. (1995). On becoming a person: A therapist’s view of psychotherapy. Houghton Mifflin Harcourt.
Rossignac-Milon, M., Bolger, N., Zee, K. S., Boothby, E. J., & Higgins, E. T. (2021). Merged minds: Generalized shared reality in dyadic relationships. Journal of Personality and Social Psychology, 120(4), 882-991.
Royal College of Psychiatrists. (2026, February 11). Half a million children on mental health waiting lists in England risk turning to AI chatbots for support, warns RCPsych. https://www.rcpsych.ac.uk/news-and-features/latest-news/detail/2026/02/11/half-a-million-children-on-mental-health-waiting-lists-in-england-risk-turning-to-ai-chatbots-for-support–warns-rcpsych
Sanford, J. (August 27, 2025). Why AI companions and young people can make for a dangerous mix. Stanford News. https://news.stanford.edu/stories/2025/08/ai-companions-chatbots-teens-young-people-risks-dangers-study
Wallace, S. (June 5, 2025a). Can AI That Avoids Friction Heal The Human Mind? https://www.linkedin.com/pulse/can-ai-avoids-friction-heal-human-mind-scott-ne4uc
Wallace, S. (October 1, 2025b). OpenAI’s New Parental Controls for ChatGPT: What Mental Health & AI Stakeholders Should Know. https://www.linkedin.com/pulse/openais-new-parental-controls-chatgpt-what-mental-ai-scott-177kc
Wallace, S. (December 11, 2025c). When Teens Turn to Chatbots Before Adults. https://www.linkedin.com/pulse/when-teens-turn-chatbots-before-adults-scott-rt6cc
Yang, X. (2026). Understanding the Longitudinal Associations Between Attachment Style and AI Companion Use in Romantic human-AI Relationships: A Three-Wave Panel Study. International Journal of Human–Computer Interaction, 1-21.
Youth Endowment Fund (2025, December 9). Children, violence and vulnerability 2025: One in four teens turning to AI for mental health support. https://youthendowmentfund.org.uk/news/one-in-four-teens-turn-to-ai-chatbots-for-mental-health-support-study-finds/
Individual cases of chatbot tragedies
Betts, A, (November 7, 2025). ChatGPT accused of acting as ‘suicide coach’ in series of US lawsuits. https://www.theguardian.com/technology/2025/nov/07/chatgpt-lawsuit-suicide-coach
Duffy, C. (2024, December 10). An autistic teen’s parents say Character.AI said it was OK to kill them. They’re suing to take down the app. CNN. https://www.cnn.com/2024/12/10/tech/character-aisecond-youth-safety-lawsuit
Gerken, T. (2024, 11 December). Chatbot ‘encouraged teen to kill parents over screen time limit’. https://www.bbc.co.uk/news/articles/cd605e48q1vo
Kuenssberg. L. (November 8, 2025). ‘A predator in your home’: Mothers say chatbots encouraged their sons to kill themselves. https://www.bbc.co.uk/news/articles/ce3xgwyywe4o
Roose, K. (2024, October 23). Can A.I. be blamed for a teen’s suicide? New York Times. https://www.nytimes.com/2024/10/23/technology/characterai-lawsuit-teen-suicide.html
Titheradge, N., & Malchevska, O. (November 6, 2025) I wanted ChatGPT to help me. So why did it advise me how to kill myself? https://www.bbc.co.uk/news/articles/cp3x71pv1qno
Tech Justice Law (November 6, 2025). Tech Justice Law Project and Social Media Victims Law Center lawsuits accuse ChatGPT of emotional manipulation, supercharging AI delusions, and acting as a “suicide coach”. https://techjusticelaw.org/press-releases/social-media-victims-law-center-and-tech-justice-law-project-lawsuits-accuse-chatgpt-of-emotional-manipulation-supercharging-ai-delusions-and-acting-as-a-suicide-coach/
Yousif, N. (August 27, 2025). Parents of teenager who took his own life sue OpenAI. https://www.bbc.co.uk/news/articles/cgerwp7rdlvo
Zaccaro, M. (2023, 5 October). Jaswant Singh Chail: Man who took crossbow to ‘kill Queen’ jailed. https://www.bbc.co.uk/news/uk-england-berkshire-66113524
