Digital health and clinical operations
Can AI Chatbots Help More Men Reach Mental Health Care?
See what chatbot studies show about mental health referrals and symptoms, and how a trial could test whether more men complete a first visit.
By Lampgen Clinical Research

3 min read · estimated
What the source reported
AI chatbots may make it easier for some men to talk about distress. Whether that first conversation helps more of them reach a mental health professional remains an open question. In her November 9 STAT story, Annalisa Merelli spoke with men who valued being able to talk on their own schedule and with less fear of judgment. A 2026 Pew Research Center survey found that 8% of U.S. men and 11% of women said they use chatbots for emotional support or advice. The interviews explain why the option appeals to some people. The survey measures use. The next question is what happens after that first conversation. [1–3]
Separate the evidence on symptom relief from the evidence on access
The 2025 Therabot trial randomized 210 U.S. adults with significant depression or anxiety symptoms or at high risk for an eating disorder. It compared four weeks of a purpose-built, expert-tuned chatbot with a waitlist. The chatbot group had greater improvements on the study’s symptom and risk measures at weeks four and eight. The study measured short-term change in a screened population with research oversight. An active comparison would help determine how much benefit the chatbot adds beyond other forms of support. [4, 5]
A 2026 randomized trial in 995 university students in Israel compared a different conversational AI tool with face-to-face group therapy and a waitlist. After 12 weeks, anxiety scores improved more in the AI group than in either comparison group. The study focused on symptom change in students with psychological distress. Its population and outcome differ from those of an access study enrolling men outside care. [6]
A 2024 observational study asked a question closer to access. Across England’s NHS services, referrals rose 15% where patients could use a self-referral chatbot and 6% at comparison services. The analysis included 129,400 patients. Site differences could contribute to the observed change. Referral volume shows how many people entered the pathway. To learn whether more patients received care, the next study should follow them through to a completed visit. [7]
Test safety in the pathway people will actually use
In a 2025 study using simulated mental health scenarios, researchers found examples of chatbots responding unsafely to indirect signs of crisis. Those simulations identify situations for study teams to test before enrollment. Use realistic scenarios to test the entire escalation path, including whether the responsible clinical team receives and acknowledges an alert after hours. In the consent materials, explain who can access chat data, how long it is retained, and whether it can be reused for model training. [8]
What this means for trial teams
Measure a completed first visit. Randomize eligible men who are outside care to the same referral pathway with or without chatbot support. Prespecify the proportion of all randomized participants who complete a first visit with a qualified mental health professional within 30 days. Use consented confirmation and track intermediate steps such as referral acceptance, scheduling, wait time, and attendance. That makes it possible to see whether the chatbot helped people enter care or shifted the bottleneck to the appointment queue.
Check who makes it through. Include a staff-assisted route, accessible language options, and participants who work irregular hours or live far from services. Track where people drop out and report results by age, language, and geography where the sample supports it. Follow symptoms and safety events beyond the first appointment to assess whether any gain in access persists and translates into benefit. [9]
Verify that urgent handoffs reach the team. In the monitoring plan, name the qualified clinical team that owns urgent response. Track the time from alert to acknowledgment and confirm that actions were documented. A study monitor can verify that the plan was followed and raise recurring gaps for the sponsor to address. [10]
Sources
- Annalisa Merelli, STAT reporting on AI chatbots and men’s mental health, November 9, 2026
- CDC/NCHS, Mental Health Treatment Among Adults: United States, 2024
- Pew Research Center, The gender gap in AI in the US, June 2026
- Heinz et al., Therabot randomized trial, NEJM AI, 2025
- Dartmouth, Therabot study design and oversight, March 2025
- Shoshani et al., randomized trial of a conversational AI agent, JAMA Network Open, 2026
- Habicht et al., NHS self-referral chatbot observational study, Nature Medicine, 2024
- Moore et al., simulated mental health safety evaluation, ACM FAccT, 2025
- Lampgen Clinical Insights, patient access and clinical trial participation burden
- Lampgen Clinical Insights, clinical trial issue closure and risk control