DooDooLamb News
Commentary: AI can be more comforting than a person. Here's what humans can learn from it
Brief published October 5, 2026 ยท Original source published October 3, 2026
Original reporting at channelnewsasia.com.
Automated brief. Verify important details at the original source.
What happened
Psychology researchers from the universities of Manchester and Durham ran five experiments comparing emotional-support messages written by people against those produced by large language models. Their finding, presented as a commentary, is that generative AI can produce support messages that are, under certain conditions, more comforting than those written by humans. The researchers argue this capability could be used to help people become better at supporting one another, framing AI as a potential tool for improving human emotional skills rather than replacing them.
Why it matters
Builders developing mental health or peer-support tools now have researcher-backed framing to consider: LLM-generated support messages may outperform human-written ones in measurable ways. That raises both design opportunities and questions about when and how to surface AI-generated responses versus human ones in sensitive contexts.
What to watch
The experiments compared message quality, but it is unclear what metrics defined comfort or how participants were selected. Independent replication and longer-term outcome data would clarify whether the effect holds outside controlled conditions.