What is MentalHealthBench?

AI safety is a field focused on preventing accidents, misuse or other harm from artificial intelligence systems. Part of that work uses benchmarks, which are sets of standard tests run against a system to judge its performance. People now turn to AI chatbots for emotional support and life advice, but, according to OpenAI, most existing tests of AI in this area focus on emergencies, which gives little signal about everyday conversations. MentalHealthBench is a test set of 1,215 realistic mental health conversations, from daily well-being topics to urgent emergencies, each paired with criteria written by licensed mental health experts for judging an AI's reply. It was published by OpenAI in September 2026.

Written from primary sources and reference pages, and checked against them by an independent AI fact-checker. Drafted with AI assistance.