If you feel low or anxious but not in crisis, you may wonder whether talking to an AI chatbot could help. One randomised trial offers a limited answer: a particular AI conversation programme was linked to better mood and less anxiety over four weeks in young adults with mild symptoms, compared with people who were waiting. That is not the same as therapy, and it says nothing about any chatbot you might find online.
The trial was published in October 2025 in the journal Applied Psychology: Health and Well-Being. The authors are based in China, and the AI agent in the study was built on a large language model and delivered through a short-video app.
A total of 865 young adults were randomly assigned to one of two groups. The intervention group had a dialogue with the AI agent for 28 days. The comparison group was a waiting group, meaning they received no intervention during the trial. Everyone completed three psychological questionnaires, which measured depression and anxiety symptoms at different points.
The trial was registered on ClinicalTrials.gov (NCT06346496). The abstract does not give details of exactly how the questionnaires were scored, how many people dropped out, or how much each person actually used the agent, so we do not state those.
Compared with the waiting group, the people who talked with the AI agent had significantly lower depression scores after two weeks. After four weeks, both depression and anxiety were significantly lower in the intervention group.
The authors conclude that this kind of conversational agent could ease mild anxiety and depressive symptoms in young adults with negative emotions, and they add that this applies when the companion bot is used enough. In other words, the benefit was linked to engagement, and a person who barely opens the app should not expect the same result.
The published summary does not give the size of the changes in numbers, so we cannot say how large or how lasting they were. The abstract reports a significant depression result at two weeks and significant depression and anxiety results at four weeks.
This is one trial of one specific AI programme. It does not show that all chatbots, including the general-purpose ones many people use, have these effects. It also does not show that an AI conversation works as well as psychotherapy: the comparison was with waiting, not with a human therapist or another treatment, so it cannot say how the two compare. A waiting group does not receive any attention or activity, so some of the improvement may reflect having a daily routine, a sense of being heard, or simply the passage of time.
Follow-up was short. We do not know whether the changes lasted after the 28 days ended. The participants were young adults with mild symptoms, so the findings do not apply to people with severe symptoms, to older adults or to children and teenagers. Symptoms were self-reported.
AI tools are not therapy and are not for crises. Other research, including a simulated safety test of AI companion bots, has found cases where bots agreed with unwise ideas, so any tool of this kind should be treated with care. If you are in distress, thinking about harming yourself, or worried about your safety, contact a doctor or other qualified professional. In Australia, Lifeline is on 13 11 14, and 000 is for emergencies.
Source: Effect of an AI agent trained on a large language model (LLM) as an intervention for depression and anxiety symptoms in young adults: A 28-day randomized controlled trial DOI: 10.1111/aphw.70067
Conceptual illustration of an AI-assisted conversation; not a depiction of a tested product or clinical result. AI-generated editorial illustration. This article summarises published research in our own words and is general information, not personal health advice.
Please Login or Sign up to be able to comment