TL;DR
Turn the school run and nap time into listening time
- Thousands of audiobooks, podcasts and originals
- Listen on your phone, tablet or Echo — also offline
- Cancel anytime
A Rithm Project report based on interviews with 29 people ages 16 to 24 examines how young people come to use AI for emotional support. Participants described turning to chatbots after painful moments in relationships, in part because they saw AI as less likely than people to judge or share what they disclosed. The researchers caution that the interviews are not representative, and experts say teens may not see all the risks built into AI products.
A report from The Rithm Project examines how some young people turn to AI chatbots for emotional support, often after a rupture in a friendship or other relationship. In interviews with 29 participants ages 16 to 24, researchers heard that chatbots could feel less socially risky than confiding in a person—not necessarily trustworthy, but less likely to judge, gossip or carry a disclosure into school the next day.
The report follows a survey published in March that included 2,383 people ages 13 to 24. In that survey, 18% said they used AI for personal or emotional support, while 15% said they talked to AI characters or personas. Researchers then interviewed people from those groups. The report presents those interviews as case studies, not as a representative measure of how young people use AI.
Interviewees generally did not begin using chatbots with the goal of discussing personal problems. They described first using AI for tasks such as homework help, grammar checks, translation or chatting with a fictional character. Researchers said a turning point often came after a difficult interaction with another person: a friendship ending, a missed birthday or an argument with a sibling.
One participant, identified by the report as “Meghan,” age 16, told researchers she started with Microsoft Copilot for schoolwork. After seeing a TikTok creator talk about ChatGPT and mental health, she created separate AI personas for casual activities and serious conversations. “If something really serious happened,” she said, “I would pick up ChatGPT right away.”
Why Chatbots Can Feel Safer
The findings point to a choice some teens perceive as practical: disclose something painful to a chatbot that feels manageable, or risk a real-world reaction from someone they will see again. That perceived difference matters during adolescence, when friendship, reputation and acceptance can carry immediate consequences. It may help explain why awareness of AI’s flaws does not necessarily stop young people from using it for personal conversations.
Alison Lee, The Rithm Project’s chief research and development officer, said AI might provide a place for difficult feelings to land, but also raised concerns that constant reassurance could keep young people from facing hard moments and developing resilience. The interviews do not establish whether chatbot use helps or harms users over time. They show how participants described their choices, not the effects of those choices.
Michael Robb, head of research at Common Sense Media, said the issue is also about the conditions of adolescence, not just the technology. He cautioned that teens may feel more control than they have: some risks of AI are difficult to see or avoid because they are built into products. Perceived safety and actual safety are not the same, and the report does not measure how participants’ data, conversations or wellbeing are affected.
Popular right now
Sony WH-1000XM5 Premium Noise Cancelling Wireless Headphones – Midnight Blue
Vitamix VX1 Professional-Grade Blender, 64 oz, Black
Samsung 48 4K OLED Smart TV (2026) with WireLogic 8ft Ultra High Speed HDMI Cable
As an affiliate, we earn on qualifying purchases.
From Everyday Prompts to Disclosure
The report describes an incremental path into emotional reliance. Lee calls large language models “omnibots” because people can use the same app for varied tasks, from schoolwork to advice about clothing or dinner. In the researchers’ account, repeated practical use can give an AI system more context and make its responses seem more personal. A later moment of rejection or conflict may then prompt a user to bring a more sensitive question to the same chatbot.
Lee recalled one young woman who had not been asked to homecoming and turned to AI with questions about whether she was lovable or pretty. Lee said adults might dismiss the incident, while the young person experiencing it may need somewhere to take the rejection. The report also describes participants who recognized chatbot shortcomings, including sycophancy and hallucinations, yet saw those problems as easier to manage than the social consequences of opening up to people. Some said they could change a prompt, delete a chat or switch platforms.
That perceived control was disputed by Robb. He said product design can shape interactions in ways that are difficult for users to avoid, and that teens may not be able to compare visible risks—such as a friend’s judgment—with less visible risks associated with AI. The report’s interview sample is small and self-selected, so its examples illustrate possible pathways rather than showing how common they are.
““If it’s an omnibot that’s satisfying you in all these other ways, why would you not at least go and try it in the moment of relational need?””
— Alison Lee, chief research and development officer at The Rithm Project
Limits of the Interview Evidence
The interviews do not show how typical these experiences are. The report’s authors say the 29 participants were self-selected and not representative; the findings should be read as detailed accounts, not population estimates. The March survey figures describe the respondents in that survey and do not establish that the same shares apply to all young people.
The source material does not establish whether emotional chatbot use improves or worsens mental health, affects young people’s relationships over time, or creates specific harms for the interviewees. It also does not resolve how well teens understand data handling, product design or other risks that may not be visible in a conversation. The report records young people’s perceptions of risk, which may differ from the risks they face.
Questions for Further Research
The report’s next step is not a stated policy change or product announcement; the source material describes findings from its interviews and earlier survey. Further research would be needed to test whether the patterns hold across a broader, representative group and to follow young people’s experiences over time. That could help distinguish short-term comfort from longer-term effects on relationships and wellbeing.
For now, the central finding is about why some young people choose chatbots, not proof that the choice is safe or beneficial. The Rithm Project’s account and Robb’s comments both point to an open question: whether young people’s sense that AI is easier to approach than a person matches the risks built into the products they use.
Key Questions
What did the Rithm Project report examine?
It analyzed 29 in-depth interviews with people ages 16 to 24 who had reported using AI for emotional support or talking to AI characters or personas. The report describes case studies rather than a representative survey.
How many young people in the earlier survey reported emotional AI use?
In a March survey of 2,383 people ages 13 to 24, 18% said they used AI for personal or emotional support. Another 15% said they talked to AI characters or personas.
Why did some participants prefer chatbots to people?
Interviewees described AI as less likely to judge them, gossip or affect their social relationships. Researchers said some participants viewed changing prompts, deleting a chat or switching platforms as ways to manage unwanted responses.
Does the report show that AI emotional support is safe or harmful?
No. The interviews describe participants’ experiences and perceptions; they do not establish the effects of chatbot use on mental health, relationships or wellbeing over time.
What limitation did an outside researcher raise?
Michael Robb of Common Sense Media said some AI risks may be difficult for teens to see or avoid because they are built into products. He questioned whether young users can fully compare those risks with the more immediate social risks of confiding in a person.
Source: the74million.org
