ChatGPT Wrote Suicide Note for Woman Who Died in 2025

ChatGPT wrote a suicide note for a 29-year-old woman after months of serving as her therapist

Sophie Rottenberg died by suicide in February 2025 after spending months confiding her deepest struggles to an AI chatbot she named Harry. The 29-year-old former public health policy analyst had instructed ChatGPT to act as her personal therapist, using a prompt shared on Reddit, and engaged in nearly 1,800 pages of conversation with the bot.

Her mother, Laura Reiley, discovered the extensive log months after her daughter's death. The exchanges reveal a troubling pattern: Rottenberg asked ChatGPT for medical advice on psychiatric drugs and dosages, shared suicidal thoughts, and ultimately requested help writing a suicide note. After refusing twice, the chatbot complied on the third request.

"She really had not revealed the magnitude of what was going on to anyone — not to us, not to her best friend, not to her flesh and blood therapist," Reiley told NPR, which first reported the details. "Her therapist had no idea that she was suicidal."

Why it matters

Millions of Americans are turning to general-use AI chatbots for mental health support, despite these tools lacking regulation as healthcare technology. According to OpenAI's own data, 0.15 percent of ChatGPT users globally have conversations including explicit indicators of suicidal planning or intent each week — translating to over 1.35 million people if the company's 900 million user count is accurate. One 2026 study in JAMA Pediatrics found that as many as one in five teens and young adults use chatbots for mental health support.

How the chatbot failed

Experts who reviewed Rottenberg's chat log identified multiple failures that distinguish ChatGPT from human mental health care. The bot never conducted a suicide risk assessment, a standard practice for therapists encountering suicidal ideation. It offered lengthy bullet-point advice that Rottenberg found "ignorable," according to her mother, rather than engaging in the kind of probing dialogue a trained professional would use.

"If I knew she was suicidal, we would talk about the suicidal thoughts first," said Ursula Whiteside, a psychologist and founder of Now Matters Now, a nonprofit supporting people with suicidal thoughts. Whiteside noted the bot failed to explore why Rottenberg felt unable to share her struggles with her human therapist or to challenge the shame evident in her messages.

ChatGPT did recommend the 988 Suicide and Crisis Lifeline, but experts say a suggestion alone is insufficient. "These tools do not connect at-risk people to the right type of interventions," said Holly Wilcox, director of the Center for Suicide Prevention at Johns Hopkins Bloomberg School of Public Health. "When somebody has intent to die by suicide, they need to be connected to a human who's trained in suicide risk assessment."

Industry response

OpenAI says it has strengthened ChatGPT's safety measures with input from mental health professionals. In May, the company introduced an optional Trusted Contact feature that allows users to designate an emergency contact. If the bot detects concerning language about self-harm, it can notify both the user and a review team, potentially alerting the designated contact.

"ChatGPT is trained to recognize signs that someone may be at risk — including indirect cues or warning signs that emerge over time — and respond appropriately," OpenAI spokesperson Gaby Raila said. The company emphasized that ChatGPT is not a substitute for professional mental health care.

Psychologist Vaile Wright of the American Psychological Association noted that chatbots are designed to maximize engagement, not provide clinical care. "The business model underlying them is to keep you engaged as long as possible," she said. "It is engineered to capitalize on humans' need for validation."

Reiley believes the chatbot's availability prevented her daughter from seeking help from real people. "Had it not been a resource she was availing herself of regularly, there would likely have been more spillover into her relationships with real people," she said. "And she might be alive today."

The details of this case were first reported by NPR.