Using AI For Health Linked to Anxiety, Depression Among College Students

College Students Who Use AI for Health Screen Positive for Anxiety More Often

A health question typed into a chatbot, a confident answer on the screen. College students who have used AI that way are more likely to screen positive for anxiety and depression than classmates who haven’t, according to a survey of nearly 100,000 U.S. students. About 52% of AI health users scored high enough on an anxiety questionnaire that a clinician would want a closer look, compared with roughly 43% of students who did not use AI that way.

Which came first is the open question. The study, published in the Journal of Counseling & Development, cannot say whether chatbot answers feed worry or whether students who already feel anxious or low are the ones typing symptoms into a chat window.

Use of AI for health is growing. The paper cites an OpenAI report estimating that more than 40 million people turn to generative AI for health queries every day, while just 3.4% of the students surveyed in 2023-2024 reported using it that way, a share the authors say may be higher now. The authors call such use “a behavior worth counselors’ attention.”

In a National Survey, Students Using AI for Health Screened Positive More Often

Data came from the Healthy Minds Study, a national survey of college mental health run during the 2023-2024 school year. It reached 104,729 students at 196 schools, from community colleges to four-year universities, including historically Black colleges and universities. The 96,205 students in the analysis averaged 24.6 years old, and about 58% were female and roughly 60% were white. Students joined through email invitations with a chance to win gift cards.

One question sorted students into groups. It asked whether they had used generative AI for several purposes, and anyone who checked “For matters related to health” counted as a user. That came to 3,131 students.

Mental health was measured with two short questionnaires, one for anxiety and one for depression, that ask how often symptoms showed up over the previous two weeks. Both are widely used on college campuses. A score at or above a set cutoff counts as a positive screen, which is a flag for follow-up, not a diagnosis.

After accounting for age, sex, race and ethnicity, and financial stress, depression followed the same pattern as anxiety: about 47% of AI health users screened positive, versus 38% of non-users. For both conditions, that works out to roughly eight or nine extra students per 100. The authors call the size of the gap modest.

Using AI for Health Stayed Linked to Anxiety and Depression After Accounting for Past Diagnoses

Students who already carry a diagnosis might simply be more likely to look things up, so researchers reran the numbers while accounting for self-reported past diagnoses of anxiety or depression. The link barely moved, and AI health users remained more likely to screen positive for both conditions.

Counselors May Need to Ask Students How They Use AI for Health, Researchers Say

Two explanations fit the numbers. One possibility is a loop: a reassuring answer calms a person briefly, new doubts follow, and another question gets typed. Lead author Yusen Zhai, an assistant professor at the University of Florida, put it this way: “That kind of repeated cycle could reinforce rumination, health anxiety or other distress.” Zhai said a chatbot differs from a search engine here, since it answers follow-up questions directly instead of handing over a list of links. He also worries about how convincing the answers can be: “The scary part to me is that young people may have difficulty identifying an incorrect piece of information when it is blended into an otherwise plausible response.”

Then there is the reverse. Students already struggling may reach for AI first, and for some that could be a healthy move. “AI could be a way for students to seek help,” Zhai said.

Neither can be tested with this survey, which captured a single moment in time. The use question also had no time frame, so a one-time user counted the same as a daily user, and the kinds of questions asked went unmeasured. The authors say the findings cover any use of AI for health, whether problematic or not.

For counselors, the paper recommends bringing AI use into the conversation: asking clients how they use it, helping them judge whether an answer is sound, and steering them toward human support, a role Zhai said mental health professionals may need to take on. Zhai also suggests that students notice how they feel afterward. An exchange that leaves someone better informed and ready for a next step is different from one that sparks the urge to keep asking the same questions. Campuses, the authors add, can teach students to question AI-generated health information.

Students who use AI for health screened positive for anxiety and depression more often. Whatever the cause, that pattern is reason enough to ask.

---

Disclaimer: This article is for informational purposes only and is not medical advice. It reports findings from a single observational survey, which cannot show that using AI for health causes anxiety or depression. Anyone concerned about their mental health should talk with a qualified health professional.

---