Chatbots match human support in everyday emotional scenarios

Research involving Durham University’s School of Education found that AI chatbots, including ChatGPT, can match humans as a source of everyday emotional support. Led by the University of Manchester, the work examined large language model responses in non-clinical situations such as being passed over for a promotion or returning to an untidy shared home.

The research comprised five studies involving a total of 1,233 participants. Messages generated by three leading LLMs, ChatGPT 4o, Claude 3.5 Sonnet, and Gemini 1.5 Pro, were compared with responses written by human participants. Recipients were not told whether a message came from a person or AI, though many correctly identified the source.

LLM responses largely reflected the strategies humans use to manage others’ emotions, with ChatGPT showing the closest similarity to human patterns. For anger and fear, LLM-written messages were rated as more emotionally supportive than human-written ones and helped participants feel more positive, such as calmer or happier. For sadness, AI and human responses were equally effective.

Simple emotional validation did not fully account for the results. Clear, actionable suggestions, including breathing techniques, step-by-step reframing, or practical next steps, improved emotional outcomes whether they came from a person or an AI.