A surprising new analysis suggests that patients may find artificial intelligence more compassionate than real doctors.

Researchers from the Universities of Nottingham and Leicester reviewed 15 studies comparing patient interactions with AI chatbots like ChatGPT to those with human healthcare providers. On a 10-point empathy scale, patients rated their chatbot experiences about two points higher than their experiences with real clinicians. Overall, AI responses had a 73% likelihood of being judged more empathetic than those from humans.

According to Study Finds, this advantage held in 13 of the 15 studies, which covered topics ranging from cancer and mental health to thyroid conditions and general medical questions. The findings challenge a report by the U.K. government claiming that empathy is an “essential human skill that AI could not replicate.”

Surprisingly, AI performed best when handling patient complaints and grievances. In those cases, ChatGPT-4 scored 2.08 standard deviations higher than human counselors. Both patients and physicians who reviewed the chatbot’s replies described them as “warm,” “understanding,” and “considerate of patient concerns.”

A chatbot is a computer program that simulates and processes human conversation through voice or text. The only specialty where human practitioners outperformed AI was dermatology.

It’s important to note that the studies were based solely on written responses, not spoken communication, and did not involve real patients. Instead, healthcare professionals, medical students, patient representatives, and researchers evaluated the AI and human responses.

While empathy is known to help reduce patient pain and anxiety, experts caution that the benefit disappears if the medical advice itself is inaccurate. The study authors suggest a hybrid approach in which doctors draft initial responses and AI systems refine the tone and empathetic language to better support patients.