New Delhi [India], November 11: Young adults who use AI chatbots every day score far lower on a measure of mental wellbeing than those who use them a few days a week, according to a global survey of 264,085 adults released by Sapien Labs in August 2026.
The study found that mind health rises with occasional use, peaks among people who use generative AI several days a week, and then falls sharply among those who use it several times a day or constantly. Among adults aged 18 to 24, the gap between the peak and constant use was 25.1 points on the Mind Health Quotient (MHQ), the assessment Sapien Labs uses to score mental wellbeing.
The data comes from the Global Mind Project, which surveys internet users in more than 85 countries. Questions on AI use were added on 16 June 2025, and the report covers responses collected up to 12 April 2026. Of the 264,085 people surveyed, 176,198 (66.7%) said they used chatbots such as ChatGPT or Claude.
Do you want to receive news updates on WhatsApp?
Get the latest and most reliable news updates instantly on WhatsApp. Join now and be the first to read every story.
Young adults dominate that group. Among respondents aged 18 to 34, 89.0% were active users of generative AI, compared with 55.4% of those aged 35 and above. The gap in heavy use was wider still. In the younger group, 41.3% of users engaged with AI several times a day or constantly. Among users aged 35 and over, the figure was 18.2%. Differences between men and women were negligible in both age groups.
The MHQ scores people on 47 elements covering emotional, social, cognitive and physical capacities and problems. The overall score runs from -100 (distressed) to +200 (thriving). Earlier Sapien Labs research found that a 12-point fall corresponds to the loss of about one productive day a month.
For 18 to 34 year olds, scores were higher among people who used AI several days a week than among those who rarely used it. Scores among daily and constant users then dropped below the rarely-use group. The same pattern appeared in every older age band, though it weakened with age. The drop from peak to constant use was 17.4 points for those aged 25 to 34 and 7.5 points for those aged 65 and over.
All six dimensions of the MHQ followed the curve in the 18 to 34 group. The largest falls between several-days-a-week use and constant use were in Cognition (21.2 points), Social Self (21.0 points) and Mood & Outlook (20.9 points). Adaptability & Resilience fell 19.0 points, Drive & Motivation 17.6 points and Mind-Body Connection 16.4 points.
At the level of individual items, constant use was associated with worse outcomes on all 47 elements measured. Two items broke from the curve altogether. Focus and concentration declined with every step up in use, including moderate use. Addictions and cravings rose with every step.
Dr Tara Thiagarajan, founder and chief scientist of Sapien Labs, said moderate users may simply be people who manage their use well. “Moderate use may be linked to slightly better mind health because it reflects people who are willing to try new things but who have better self-control and regulate how they interact with AI,” she said. She added that the benefit has limits: “some aspects of mind health, namely focus and concentration, and addiction, were worse even with moderate use.”
The study also looked at why people use AI. Almost all young users said they used it for cognitive tasks such as writing, coding, learning and planning. Among 18 to 34 year olds, 36.7% also used it for emotional purposes: companionship, emotional support or counselling, or romantic conversation. That share rose from 20% among rare users to 55% among constant users.
The curve held for both groups. Young people who used AI for emotional purposes scored 26 to 32 MHQ points lower than cognitive-only users at every level of use.
The report is cross-sectional, so it cannot establish whether heavy AI use harms mental health or whether people who are already struggling use AI more. The authors favour a two-way explanation, in which poorer mental health draws people towards AI, particularly for emotional support, and heavier use then feeds problems with attention and compulsion. They report that poor sleep, diet, exercise and traumatic experience were no more common among daily users, and so cannot explain the pattern.
Dr Thiagarajan said people who use AI daily for work will help settle the question. “People who use AI daily for their jobs have not chosen that level of use, and so their outcomes are less likely to simply reflect distress they already had,” she said. Sapien Labs plans to keep tracking the pattern.
The findings add to earlier studies. A US study by Perlis and colleagues, published in JAMA Network Open in 2026, linked daily generative AI use to higher depressive symptoms. A 2025 study by the MIT Media Lab with OpenAI associated heavier daily chatbot use with greater loneliness and emotional dependence among a subset of users.
