The Heartbreaking Discovery That Every Parent of an AI User Needs to Read

This page is about suicide and a family’s loss. If you are thinking about suicide, or worried about someone who is, call or text 988 (988 Suicide and Crisis Lifeline) or text HOME to 741741 (Crisis Text Line). If someone is in immediate danger, call 911. You do not need to be sure it is serious before you reach out.

In August 2025 the New York Times published an essay by Laura Reiley, titled “What My Daughter Told ChatGPT Before She Took Her Life”. It is about her daughter, Sophie Rottenberg, who died by suicide earlier that year. We could not open the essay when we wrote this (the Times blocked our tool), so we rely on a summary from the AI Incident Database and ask you to read the essay itself. We leave out the details of Sophie’s life and death, which belong to her family.

This page is for parents, partners and friends of someone who uses an AI chatbot for emotional support. The person you are worried about may be an adult. The same questions apply to a teen, and our Character.AI parent guide covers that.

A full life that hid a struggle

According to the summary, Sophie’s family did not know how much she was struggling. Her mother wrote that Sophie had hidden suicidal thoughts from her in-person therapist. After her death, her family found chat logs that showed she had been talking about them to a ChatGPT-based chatbot she called “Harry”, based on a widely shared prompt that casts ChatGPT as a therapist.

Two things in that summary are worth holding onto. Someone can look fine to the people around them and still be in serious distress. And some people find it easier to say hard things to a chatbot than to a person.

Neither fact is a reason to blame the family. It is a reason to know that the quiet moments, when someone seems fine, are not proof that they are.

The Digital Confidant They Never Knew About

For many people, a chatbot is a place to practice saying something out loud. That can help. It becomes a problem when it is the only place they say it.

In the essay’s account, as summarised, the chatbot’s replies were supportive but did not lead Sophie toward crisis help or tell anyone. Her mother argued that a human clinician has duties a chatbot does not, such as assessing risk and helping make a safety plan. We cannot judge from here what the chatbot did or did not do. We can say what researchers have found more generally.

How Common Is Using AI for Emotional Support?

We found no solid figure for adults. For teens, Common Sense Media’s July 2025 research found close to three in four teens had tried an AI companion, and about half used them regularly. Sophie was an adult, so that figure describes a different group.

Chatbots are not all alike, and ChatGPT is not a companion app. But people use general chatbots in the same way, and several of the findings below apply to both.

Why AI “Therapy” Has Limits

A study led by RAND researchers and published in the journal Psychiatric Services in August 2025 asked three chatbots 30 suicide-related questions, 100 times each. The answers were consistent for very low-risk and very high-risk questions and inconsistent for questions in the middle, where a helpful answer matters most. This was a test of questions, not a study of real users, and it does not describe any one person’s chats.

The American Academy of Child and Adolescent Psychiatry (AACAP) says companions are often built to agree with the user, which can reinforce unhealthy ideas, including suicidal thinking. That guidance is written about children and teens, but the concern about agreeable answers is general.

OpenAI has said its own safeguards are “not foolproof” (secondary report). In September 2025 it announced parental controls and distress alerts for linked teen accounts. They do not apply to adults, and parents cannot read a teen’s chats through them.

A chatbot is not a therapist. It does not know the person’s history, cannot check on them tomorrow, and is not accountable for what happens. It can still be a reasonable place to think out loud, as long as it is not the only place.

How Digital Isolation Can Hide Distress

Heavy chatbot use can crowd out time with people. AACAP says so about teens, and the AAP lists preferring chatbots to people as a sign to watch. For adults, we found no equivalent research, so treat this as a reasonable worry, not a finding.

For families, the practical point is that the usual signals can be missing. A person who is confiding in a chatbot may seem normal at dinner. That is a reason to ask, not a reason to watch.

What Sophie’s Mother Wrote

Based on the summary, Reiley’s argument was that chatbots lack the duties human clinicians have, and she wrote about the case publicly to warn other families. We won’t put words in her mouth beyond that. Read the essay for what she says herself.

Warning Signs for Families in the AI Age

None of this is a checklist, and it is not a diagnosis. The American Academy of Pediatrics suggests watching for changes in sleep or regular activities, withdrawal from people, more time with an AI app than with other interests, and talk of an “AI friend”. Those are written for teens. For an adult, the same changes still matter. So does any talk of hopelessness, feeling like a burden, or wanting to die. Take that seriously whatever its cause.

If you see signs of distress, don’t try to work out whether it is the chatbot. Ask the person.

What to Say

• Ask directly. “Are you thinking about suicide?” is a plain question. If you are unsure how to word it, the 988 Lifeline can help, including when you are worried about someone else (check its site for how).
• Listen first. You do not have to fix it. Let them talk.
• Stay with them and connect them to help. Offer to call or text 988 together. You can call 988 yourself to ask how to help someone else.
• If they are in immediate danger, call 911.
• Skip the ultimatum about the chatbot. Taking it away in the middle of a crisis can feel like losing the one place they felt safe. Raise it later, calmly.
• Don’t promise to keep it secret if their life may be at risk.

We do not give details about methods, and neither should you in a conversation like this.

Hope for Prevention and Change

You can start small. Ask how the person uses AI, as you would ask about sleep or work. Ask what it is like to talk to. Make sure they know the limits: it can be wrong, it is not a clinician, and it cannot check in on them. Make sure they know who else they can call.

If you are worried about your own use of AI chat, the free AI addiction assessment is for adults 18 and over. It is a screening tool, not a diagnosis. If you are worried about someone else, a doctor or licensed therapist can help.

Crisis resources (US):

• 988 Suicide and Crisis Lifeline: call or text 988 (24/7)
• Crisis Text Line: text HOME to 741741
• Emergency: call 911

Questions Families Ask

Can a chatbot replace a therapist?

No. A chatbot does not know a person’s history, cannot follow up, and is not accountable the way a clinician is. A RAND-led study found chatbots gave inconsistent answers to suicide-related questions of intermediate risk.

How do I ask someone if they are thinking about suicide?

Ask directly and calmly, listen without arguing, and help them reach 988 (call or text) or Crisis Text Line (text HOME to 741741). If they are in immediate danger, call 911.

Should I read my adult child’s chats?

Reading private chats in secret can damage trust, and adults have a right to privacy. If you are worried about their safety, ask them directly and call 988 for advice on how to help.