Do AI mental health chatbots affect attention, sleep, or emotional regulation?

Do AI chatbots actually improve emotional regulation?

Yes, the strongest evidence shows small-to-moderate improvements in emotional regulation, specifically for depression, anxiety, and stress. A 2025 systematic review and meta-analysis of 31 randomized controlled trials (RCTs) involving nearly 30,000 adolescents and young adults found that AI chatbots produced significant reductions in depressive symptoms (effect size -0.43), anxiety (-0.37), stress (-0.41), and even psychosomatic symptoms like headaches or stomachaches (-0.48) [1]. These effect sizes mean that, on average, users experienced a noticeable decrease in distress compared to control groups. The same review also found modest improvements in life satisfaction and well-being, though effects on positive affect and self-efficacy were limited [1].

A separate 2025 RCT of a generative AI chatbot called Therabot, conducted with 210 adults who had clinically significant symptoms of major depressive disorder, generalized anxiety disorder, or eating disorder risk, also reported symptom improvement over 4 to 8 weeks [4]. This trial is notable because it tested a newer, more flexible generative AI system, rather than older rule-based chatbots. Together, these two large, well-designed studies [1][4] converge on the same conclusion: AI chatbots can help people manage emotional distress, at least in the short term.

What about attention and sleep? Is there any evidence?

None of the five studies directly measured whether AI chatbots affect attention or sleep. The meta-analysis [1] focused on mental distress (depression, anxiety, stress) and health behaviors, but did not include sleep quality or attention as outcomes. The postpartum chatbot trial [2] measured depression and anxiety only. The ethical review [3] and the qualitative interview study [5] discussed user experiences and safety concerns, but neither collected data on sleep or attention. This is a clear gap in the research. It is possible that improvements in emotional regulation could indirectly benefit sleep (e.g., less anxiety at bedtime), but that is speculation, not evidence. For now, if your primary concern is attention or sleep, the research does not provide an answer.

What are the important caveats and risks?

The evidence comes with important limitations. First, the effects are small-to-moderate, not large — meaning chatbots are not a replacement for human therapy, especially for severe conditions. The meta-analysis [1] noted that user engagement was critical; repetitive content and technical issues often caused people to stop using the chatbots, which reduced effectiveness. Second, a 2025 scoping review of 101 articles on ethical challenges [3] highlighted serious concerns: safety in crisis situations (e.g., suicidal ideation), the risk of users becoming dependent on the chatbot, and the problem of 'black box' algorithms where users cannot understand how the chatbot arrives at its suggestions. Third, the postpartum trial [2] found that in a general (non-depressed) population, the chatbot did not significantly reduce depression scores, suggesting these tools may be most useful for people who already have some symptoms. Finally, the qualitative study [5] reported that users valued the 'emotional sanctuary' chatbots provided, but also emphasized the need for better safety guardrails and human-like memory — features current chatbots lack.

About These Sources

This answer is built on 5 peer-reviewed studies — published from 2023 to 2025, 4 from 2024 or later, 3 in Q1 journals, collectively cited 348 times — selected as the most relevant from 5 studies that passed quality screening, drawn from 59 papers retrieved from a database of over 500 million.

Sources used in this answer

Related research questions