Artificial intelligence can make online medical articles much more readable for laypeople – but correctness remains a critical point. This is shown by a new study by Heilbronn University of Applied Sciences (HHN), which has been published in the journal JMIR AI .

The team led by Amela Miftaroski (Bachelor’s graduate in Medical Informatics), Dr. Richard Zowalla, Martin Wiesner and Dr. Monika Pobiruchin analyzed 60 original texts on common diseases and health topics. These were automatically simplified with four major language models (including ChatGPT-3.5, Microsoft Copilot) and then evaluated with established readability indexes.

Results:

• All tested models improved readability.
• Microsoft Copilot achieved the strongest improvements, reaching the recommended intermediate level (comparable to 8th-9th grade) for half of the texts.
• ChatGPT-3.5 followed closely behind, with other models showing only minor effects.
• However, the level recommended by experts (approx. 8th grade) was rarely fully reached.

Risks and limitations:

• Some AI texts contained inaccuracies, omitted important contextual information, or generated potentially misleading wording.
• “The method is therefore not suitable for private use,” emphasizes first author Amela Miftaroski. “A professional examination by doctors remains mandatory.”

Potential for practice:
The authors see great potential if AI is used as a design aid: healthcare facilities could have the first simplified versions created, which are then corrected and finalized by specialists. In the long term, this would make it possible to create more understandable health information for the population.

The study was carried out as part of a bachelor’s thesis and underlines the close interlocking of research and teaching at HHN. It shows how students can access scientific publications in international journals at an early stage.

Original Paper:

JMIR AI – Leveraging Large Language Models to Improve the Readability of German Online Medical Texts: Evaluation Study

Editor: X-Press Journalistenbüro GbR

Gender Notice. The personal designations used in this text always refer equally to female, male and diverse persons. Double/triple naming and gendered designations are used for better readability. ected.