A study shows that answers given to chatgpt patients were more appreciated than those of therapists

A new study suggests that the answers provided by Chatgpt in psychotherapy scenarios are often better evaluated than those written by therapists. The researchers found that the participants had difficulty distinguishing between the answers generated by artificial intelligence (Ia) and those written by specialists. In addition, ChatGPT responses were generally longer and detailed, thus offering a more detailed contextualization.

Artificial intelligence could revolutionize couple therapy. Photo: Shutterstock

This contextualization could explain the higher scores obtained by Ia regarding the essential principles of psychotherapy. The discoveries emphasize the potential of artificial intelligence in therapeutic interventions, but also raises ethical and practical questions regarding its integration in the field of mental health. The researchers draw attention to the need for professionals in the field to be actively involved in the development of Ia to ensure responsible supervision.

Comparing the answers given by psychotherapists with those generated by Chatgpt, the study published on February 12, 2025 in Plos Mental Health magazine shows that the latter were generally evaluated. The research was carried out by H. Dorian Hatch, from the Ohio State University, co -founder of Hatch Data and Mental Health, together with his team.

In the new study, attended by over 800 people, hatch and colleagues found that, although certain linguistic style differences were observed, the participants rarely managed to identify whether an answer came from a human therapist or by at Chatgpt in the 18 couple therapy scenarios analyzed.

This discovery confirms a prediction that people will not be able to make the difference between the answers offered by a machine and those written by a man. Moreover, ChatGPT responses have been generally evaluated better in terms of fundamental principles of psychotherapy.

An additional analysis showed that the responses generated by Chatgpt were, on average, longer than those of the therapists. Even after taking into account the length of the answers, it continued to use more nouns and adjectives. Given that nouns are used to describe people, places and objects, and the adjectives add context and details, this aspect suggests that Chatgpt manages to provide a broader contextualization than therapists.

According to authors, these results can be an early indication that Chatgpt has the potential to improve psychotherapeutic processes. In particular, this work can lead to the development of different methods of testing and creating psychotherapeutic interventions.