GAO reports · Published
Science & Tech Spotlight: AI Chatbots and Mental Health
Record date · Nov 20, 2026
Open official recordGAO report excerpt
Why This Matters Nearly one-fifth of Americans aged 12 to 21 reported using an AI chatbot for mental health advice in 2025. Experts warn that general-purpose chatbots are not designed for such needs and can be dangerous if used for this purpose. Further, while purpose-built chatbots may be designed to help, they are not a replacement for qualified mental health providers. Key Takeaways The Food and Drug Administration (FDA) has not authorized any AI chatbot to support mental health. Settings for how agreeably chatbots interact with users are critical, and some guardrails may not be effective. Challenges include the potential for harm, unproven effectiveness, and evolving regulations. The Technology What is it? AI chatbots are computer programs that simulate conversations. General-purpose chatbots allow users to ask about and receive responses on a wide range of topics. However, chatbots tend to be agreeable and flattering (sycophantic) and reinforce the user’s perspective rather than challenge it. There is no universal off switch; reducing sycophancy requires users to leverage custom settings or prompts and companies to adjust system prompts, reward signals, context framing, and other settings. Purpose-built chatbots are developed for a specific task. For example, chatbots may be designed to provide education on mental health topics, instructions for breathing or therapeutic exercises, or support between counselor or therapist visits. How does it work? Modern chatbots feature user-friendly interfaces that rely on an underlying generative AI, specifically large language models, which are trained on massive amounts of data to “learn” human language patterns. General-purpose chatbots construct their responses to users’ questions from online information and training data. Purpose-built chatbots are trained on mental health data and therapeutic techniques, according to developers. While both types of chatbots can generate confident-sounding responses based on patterns in
Record details
- GAO product
- GAO-27-109184
Sources and related records
Source checked .