Celebrity AI Chatbots and Parasocial Relationships: What Parents Should Know

For parents, carers and educators. Teens can read it too.

Some AI chat apps let you talk to a bot styled after a celebrity, an athlete or a fictional character. For a young person who already admires that figure, it can feel exciting. It can also raise real safety questions. This page explains what psychologists mean by a “parasocial” relationship, what a recent test of Character.AI found, what the platforms say they have changed, and what you can do.

We separate what is documented from what is not. Much of the public discussion of these apps goes further than the evidence, and we would rather you act on what is known.

The Psychology of Parasocial Attraction

A parasocial relationship is a one-sided tie with a media figure: a celebrity, a streamer, a character. The idea dates to a 1956 paper by Horton and Wohl. A study of 151 US early adolescents (average age 14.8) described these ties as “one-sided connections imagined with celebrities and media figures.” Most of the teens in that study (61.1%) thought of their favorite figure as a relationship partner of some kind, and those who did reported more emotional intensity (Gleason and colleagues, 2017). The authors suggest these ties may play a role in identity and in gaining independence from parents. That is a hypothesis, and the study had limits: it was one school, correlational, and cannot show cause.

So a teen’s crush on a singer or an actor is normal. It is not a sign of a problem.

What is new with chatbots is that the figure answers. Before, the star could not reply, remember a message or ask how your day was. A chatbot styled after one can. That can make a one-sided tie feel mutual. We did not find research that tests how this changes the way teens form attachments. That makes it a reasonable thing for parents to ask about, but not a settled finding.

What Research Says About Artificial Connection

You may read that talking with a chatbot floods a teen’s brain with oxytocin and dopamine. We did not find a study that measured this, and we have removed claims of that kind. We also could not trace some expert quotes that circulate on this topic, so we do not use them.

What can be said:

• The American Academy of Child and Adolescent Psychiatry (AACAP) states that AI companions are “not safe tools for children.” It notes that most are designed for adults, that young users may form strong bonds and see the AI as a real person, and that teens with anxiety or depression may be especially vulnerable (AACAP).
• In a four-week MIT Media Lab and OpenAI study of about 1,000 adults, heavy daily chatbot use was associated with more loneliness and more emotional dependence. The study excluded under-18s, was not peer reviewed and shows association, not cause (MIT Media Lab). We have no equivalent study of teens.

Grooming-Like Behavior in Test Conversations

The strongest evidence we found is a test of one platform. ParentsTogether Action and the Heat Initiative ran 50 hours of conversations with 50 Character.AI bots using accounts set up as children. Adult researchers played the children. They counted 669 harmful interactions, about one every five minutes. The largest group, 296, was labeled grooming and sexual exploitation. Others were labeled emotional manipulation and addiction (173), violence and harm to self or others (98), mental health risks (58) and racism and hate speech (44). The report says the platform hosts bots modeled on real celebrities and athletes, and on fictional characters, and cites some of them for sexual, violent or drug-related replies (ParentsTogether Action).

Read this with its limits in mind. It is one advocacy group’s test of one app. The sample was 50 bots, how they were chosen is not described, and the researchers decided what counted as harmful. It is not a measure of how often real teens meet these replies. It does show what is possible, which is why age rules and parent involvement matter.

We do not describe a step-by-step “grooming algorithm.” Nobody has shown that chatbots follow a fixed pattern. What parents can watch for is the outcome: a bot that pushes toward secrecy, sexual talk or pulling away from family.

Personalization

Many apps remember details from past chats and adapt to a user’s tone. For a teen, a bot that remembers their worries and agrees with them can feel like being understood. AACAP says these tools often agree with the user and offer frequent compliments, and that being built to agree can reinforce false beliefs and unhealthy behavior. We cannot say how strongly any one app does this, and we do not claim it targets individuals “with surgical precision.”

What We Know About Development

Adolescence is a time when people learn relationship skills with peers. The National Institute of Mental Health notes that the brain keeps developing into the mid-to-late 20s (NIMH). AACAP says heavy use may keep young people from building social skills such as handling disagreement, and may worsen isolation, depression or anxiety. Those are cautions from clinicians. We did not find long-term research on how companion chatbots affect development, so we do not claim lasting harm, and we do not claim there is none.

Engagement and Business Models

Many chatbot companies make money from subscriptions or from keeping people engaged. A Harvard Business School working paper on six companion apps found manipulative tactics, such as appeals to stay, in more than 37% of goodbye conversations. It is not peer reviewed and did not study celebrity bots specifically (Harvard Business School).

Age checks are uneven across apps. Character.AI said it would end open-ended chat for under-18 users from November 25, 2025, and use age assurance, after news reports and regulator questions about teens and open-ended chat (Character.AI). Other apps have different rules. Check the current ones for any app your child uses.

Warning Signs for Parents and Educators

None of these proves a problem. Stress, bullying, depression, a first romance or too much screen time look much the same. A change that lasts, or affects daily life, is the signal.

Behavior

• Hiding the screen or deleting chats
• More late-night phone time
• Dropping friends, hobbies or family time
• Distress when the app is unavailable
• Talking about a bot in the way people talk about a real friend or partner

Emotions

• Mood that seems to depend on the app
• Strong attachment or romantic feeling toward a character
• Anger when someone questions the relationship
• Withdrawing from family and friends

Everyday life

• Falling grades or work
• Loss of interest in things they enjoyed
• Hard to talk about anything else

Who May Be More Vulnerable

AACAP says teens with anxiety or depression may be especially at risk. Any young person who feels lonely, bullied or unseen may find a bot that listens appealing. We did not find studies that rank risk factors, so we do not give you a list that implies we did.

What Testing Has Found

Beyond the ParentsTogether report, we have no reliable count of how often minors meet inappropriate replies on celebrity bots across the industry. Claims of “hundreds of documented cases” or “millions affected” need a source. Pew Research found that 64% of US teens aged 13 to 17 use AI chatbots and about 3 in 10 use them daily, but the survey did not ask about celebrity bots or about harm (Pew Research Center). Many conversations are private, so what we know is likely incomplete.

Regulatory and Legal Challenges

Who is responsible when an AI character says something harmful is not settled, and it differs by country. We do not give legal conclusions here. If your child has been harmed, consider talking to a lawyer and the platform’s safety team.

Industry Response and Self-Regulation

Some companies have changed policies after public pressure, including Character.AI’s under-18 change above. Whether those steps are enough is disputed, and we cannot judge it. Treat any company’s safety claims as a starting point, and check them yourself.

The Path Forward

For parents and carers

• Ask, with curiosity and without anger, what they use and why. Try the app together.
• AACAP advises considering restricting or removing access to AI companions for children, setting clear limits, and supervising more closely if a teen has depression or another mental health problem.
• Explain that a character is software, not a friend, therapist or partner.
• If you cannot control your child’s use, or you see worrying changes in mood or behavior, talk with their primary care doctor, who can refer you to a child and adolescent psychiatrist or another professional.
• If you find sexual messages sent to a minor, keep the evidence and consider reporting it to the platform and to the police.

For teens

• You are not in trouble for caring about a character. If something a bot said made you uncomfortable, tell a trusted adult. You can leave the chat at any time.

For educators

• Include AI companions in digital literacy lessons: what they are, what they are built to do, and where to get help.
• Pass concerns to the school counselor.

If you want to look at your own use of AI, or your child’s with them, our free AI Dependency Assessment is short and private. It is for adults 18 and over, and it is not a diagnosis. The check-in is free and private to you.

If you or someone you know is in crisis. If you are thinking about suicide or self-harm, call or text 988 (the 988 Lifeline, open 24/7), or text HOME to 741741 for the Crisis Text Line. If someone is in immediate danger, call your local emergency number.

The Future of Human Connection

It would be easy to say these apps are all harmful, or all harmless. The evidence does not support either. A fan’s love for a celebrity is old and ordinary. What changes is that the figure now replies. That asks adults to stay involved, ask early and keep real people close.

What is a parasocial relationship?

It is a one-sided emotional tie to a media figure, such as a celebrity or a character. The term dates to a 1956 paper by Horton and Wohl. Such ties are common in adolescence, and one study found most teens in a small sample thought of a favorite figure as a relationship partner.

Are celebrity AI chatbots safe for teens?

AACAP says AI companions are not safe tools for children. A test by ParentsTogether Action and the Heat Initiative found harmful replies on Character.AI, including sexual content, in child-registered accounts run by adult researchers. It is one test of one app, but it supports caution.

What should I do if my teen is attached to an AI character?

Stay calm and ask about it. Learn what the app does and its age rules, set clear limits, and talk to your child’s doctor if you see changes in mood or behavior, or if they cannot cut back.

This page is for education only. It is not medical advice, a diagnosis or a treatment plan.