AI chatbots such as ChatGPT can match – and in some situations outperform – humans as a source of everyday emotional support, new research led by The University of Manchester and in collaboration with Durham University has found.
However, the advantage depends on the emotional context, and a key ingredient for effective support – offering specific, actionable guidance – can boost the quality of support from both humans and AI alike.
The study – published in the journal Emotion (insert date) – compared messages of emotional support generated by large language models (LLMs) and humans in everyday, non-clinical contexts.
The research comprised five studies involving a total of 1,233 participants. Across the studies, the team compared emotional support messages generated by three leading LLMs – ChatGPT 4o, Claude 3.5 Sonnet, and Gemini 1.5 Pro – with messages written by human participants responding to everyday emotionally charged scenarios, such as being passed over for a promotion or coming home to find their shared house untidy.
Participants were not told if the messages they received came from a human or an AI. Although many were able to correctly identify the source, those in studies where LLM messages were rated as more supportive did not show reduced benefit – suggesting that awareness of an AI source does not necessarily undermine its effectiveness.
Key findings:
LLM-generated responses largely mirrored strategies used by humans to manage others’ emotions, particularly for ChatGPT, which showed the highest similarity with human patterns.
In situations involving anger and fear, LLM responses were seen as more emotionally supportive than those written by people. In addition, LLM messages helped people feel more positive (such as feeling calm or happy). For sadness situations, LLM and human responses worked equally well.
Simply acknowledging someone’s feelings (e.g., “I totally understand why you feel sad”) did not fully explain why LLM responses were sometimes more effective. Adding or removing this emotional validation did not make a difference to how helpful the messages were.
What mattered most was “Actionable support” – clear practical suggestions that people could use (such as breathing techniques, step-by-step reframing, or practical next steps). These specific suggestions improved emotional outcomes regardless of whether the message came from a human or an AI.
Co-author Dr Belén López-Pérez, Lecturer in Psychology at The University of Manchester, said: “Our findings suggest that what makes emotional support effective is less about who provides it, and more about how it is delivered. Providing people with specific, actionable steps, rather than general reassurance, is a key ingredient for effective support, whether that comes from a person or a chatbot.
She added: “While a human might typically suggest going for a walk or making a cup of tea, an AI chatbot tends to go further, for example acknowledging the person’s feelings and offering step-by-step strategies such as breathing techniques or specific ways to reframe a thought. Training human supporters – whether professionals or members of the public – to provide this kind of structured, actionable guidance could meaningfully improve the quality of everyday emotional support.”
Co-author Yuhui Chen, a PhD researcher at The University of Manchester, said: “We found that AI can emulate the supportive strategies that humans use, and at times even outperform human supporters in improving a person’s emotional state – though this is not universal. The effect depends on the type of emotion and what outcome you are looking at.”
She added: “We are not saying that humans should or can be replaced by AI. But our findings do show that when people choose to use AI for emotional support, it can be genuinely helpful, and understanding why it works can help us improve how both humans and AI provide that support.”
The research comes amid growing public interest in AI as an emotional resource. Recent surveys suggest that 66% of 25- to 34-year-olds in the UK have turned to AI chatbots instead of loved ones to discuss emotional problems, and in the US, 20% of young adults aged 18-21 have used ChatGPT to manage their emotions, with 92.7% of those users finding the advice helpful.
Dr Sarah Walker, co-author of the study from Durham University’s School of Education, said: “What’s key here is that we’re not suggesting AI should replace human support. But when people do turn to it, it can help, and understanding why tells us something useful. Concrete, actionable steps work better than general reassurance, and that is something all of us can get better at offering. However, we also need to think about the human cost of actionable support as well.
“When AI suggests a walk or a breathing exercise, the suggestion is where its involvement ends. When a person offers the same thing, it usually comes with their presence as well. We don’t just suggest the walk, we go on it. That kind of support takes real time and effort, and the people closest to us are often stretched thin. So to me, this work is less about AI replacing us and more of a reminder that what we want from each other in hard moments is often more than the people we love can give.”
- “A Digital Shoulder to Cry On: Understanding Why Large Language Models Can Be Effective in Extrinsic Interpersonal Emotion Regulation is published in the journal Emotion. DOI
