Comparative Evaluation of ChatGPT, Gemini, and DeepSeek for ADHD-Related Health Information

This study systematically compared how three large language models (ChatGPT GPT-4o, Gemini, and DeepSeek R1) respond to 22 commonly asked ADHD questions across four domains, finding all models achieved high accuracy (87-91%) as independently verified by psychiatry specialists. Reproducibility was strong across models, with ChatGPT showing slightly superior performance in treatment domains while Gemini and DeepSeek excelled in basic knowledge and diagnosis.