For patients, ChatGPT is more understanding than doctors. The study

In 2019 a report commissioned by the British government stated that “empathy and compassion” remain essential human skills thatartificial intelligence he can’t reply. Six years later, a meta-analysis just published on British Medical Bulletin suggests the opposite: when they respond in writing to patients’ questions, chatbots like ChatGPT get empathy scores about two points higheron a scale of one to ten, compared to doctors, nurses and other healthcare workers. Which corresponds, in statistical terms, to a 73% probability of being judged more empathetic in a direct comparison.

Researchers from the Universities of Nottingham and Leicester examined 15 studies published between 2023 and 2024, pooling data from 13 of them. Patient questions covered oncology, thyroid disease, mental health, autism, multiple sclerosis, systemic lupus erythematosus. The largest dataset included over 2,100 interactions in a Chinese clinic, but questions also came from forums such as Reddit, emails, hospital portals. Evaluating the answers – without knowing whether they came from an algorithm or a person – were doctors, medical students, patient representatives, psychologists. “In text-only interactions, chatbots are frequently perceived as more empathetic of human healthcare workers,” the study authors write.

The patience of AI

—

The most marked difference is in the management of complaints: When it comes to responding to patient complaints, ChatGPT-4 scores 2.08 standard deviations higher than public relations workers. For thyroid questions, the difference was 1.42 standard deviations compared to surgeons; for mental health, by 0.97 compared to licensed professionals. Dermatology is an exception: in two studies on skin conditions, dermatologists outperformed chatbots, without the researchers finding an explanation for this anomaly.

From email to phone

—

All studies analyzed written communications. A detail that matters: the nod, the bending forward, the doctor’s direct gaze convey empathy as much as words, and in-person visits remain the norm. But the use of email and messaging is growing with health portals and telemedicineand one in five British doctors already use ChatGPT for tasks such as drafting emails to patients. And over 117,000 patients across 31 mental health services across the UK healthcare system have interacted with Wysa, an AI-powered digital therapist.

The study’s authors propose a collaborative model: doctors would write the initial draft, AI would improve the tone and empathetic language, then clinicians would verify medical accuracy. “It is possible that gains in perceived warmth will vanish if responses contain errors or incomplete guidance,” the researchers warn. Who are already looking at the next step: telephone consultations represent 26% of appointments with GPs in the United Kingdom and new voice systems based on artificial intelligence already promise to “respond with emotion” and “pick up on non-verbal cues”. But no study has yet put them to the test.