TUD Researchers Demonstrate the Benefits of AI-Generated Summary Reports for Doctor-Patient Consultations

Researchers at the Department of Neurosurgery at the Faculty of Medicine of the TUD Dresden University of Technology and at the Carl Gustav Carus University Hospital in Dresden conducted a randomized controlled trial to study the effects of AI-generated, layperson-friendly medical record summaries on doctor-patient consultations. As a result, patients felt better informed and more involved in decision-making; at the same time, the consultations were, on average, shorter. The results were published on October 15, 2026, in the journal NEJM AI.

Many patients have difficulty understanding doctor’s letters, test results, and radiology reports. In Germany, approximately 60 percent of adults report difficulty understanding and using health information. At the same time, there is often limited time in everyday outpatient care to explain complex medical issues in detail. Therefore, the AI-INFOCARE study, led by Dr. Elida Benker-Hasani, investigated whether presenting information in an understandable way helps patients better prepare for an initial neurosurgical consultation.

Between May 2025 and January 2026, 300 patients and 17 physicians from the Department of Neurosurgery in Dresden participated in the study. Patients were randomly assigned to receive either standard care or a summary of their medical records in plain language, generated using artificial intelligence, prior to the consultation. These summaries were created using software from Simply Onno, a startup that participated in the study as a collaborating partner.

The results show that individuals who received a summary rated the quality of the doctor-patient interaction more positively. They felt better informed and more involved in decision-making during the consultation. Meanwhile, the median duration of initial consultations decreased from 33 to 25 minutes without compromising the perceived quality of information provided.

From the physicians’ perspective, a more nuanced picture emerged. Physicians tended to rate the conversations as slightly more demanding. However, this difference was small and not statistically significant. The authors cautiously attributed this shift to a change in consultation dynamics, where explaining medical fundamentals took a back seat to joint decision-making.

"It used to be Dr. Google, but today it’s Dr. AI. Patients come into the exam room informed but not necessarily correctly informed. We wanted to know how the conversation changes when patients understand their own test results beforehand,” explains Dr. Witold Polanski, the study’s last author and deputy director of the Clinic and Polyclinic for Neurosurgery. Such summaries cannot replace doctors. However, patients who understand their test results can participate more effectively in decision-making. That is exactly what we investigated, and the consultations did not get longer but shorter.

At the same time, the researchers emphasize that these results cannot be easily generalized to other medical specialties or care settings. They did not examine long-term clinical implications, such as the impact on subsequent patient behavior or the sustainability of the effects.

Thus, AI-INFOCARE provides a building block for further research into AI in medicine, not as a substitute for medical expertise, but rather as a supportive technology whose use should be investigated step by step.

Paper: DOI: 10.1056/AIoa2600136

Note on Conflict of Interest
Dr. Witold Polanski is a co-founder and medical director of Simply Onno, the company whose software was used in the study. He was not involved in the patient interviews, data collection, or statistical analysis.

Scientific Contact:
Dr. Witold Polanski
Department of Neurosurgery
Faculty of Medicine and
Carl Gustav Carus University Hospital
at TUD Dresden University of Technology