Publication:

A cross-sectional study to assess and compare artificial intelligence-generated content for medical professionals with UpToDate on management of acute coronary syndrome

Background The rapid rise of artificial intelligence (AI) in medicine offers new ways to generate educational content, but its readability and accessibility for both clinicians and patients remain underexplored. This study compares AI-generated medical information to a standard clinical reference.

Methods This cross-sectional study analyzed educational content on acute coronary syndrome (ACS) management generated by DeepSeekAI and retrieved from UpToDate. Readability was assessed using four established metrics: Flesch reading ease (FRE), Flesch-Kincaid grade level (FKGL), SMOG index, and difficult word percentage. Data were collected across six distinct topics related to ACS management, and the scores for each tool were statistically compared.

Results Readability analysis revealed notable disparities between the two sources. UpToDate consistently produced content with significantly higher FRE scores (e.g., UpToDate FRE up to 40.2 vs. DeepSeek AI FRE down to 2.7 for topic 1) and considerably lower FKGLs, SMOG Indices, and percentages of difficult words across all examined topics. While both sources generated text with relatively high reading levels indicative of medical content, DeepSeekAI consistently demonstrated lower readability, suggesting more complex vocabulary and sentence structures.

Conclusions UpToDate content on ACS management is more readable and accessible for clinicians than DeepSeekAI-generated content, highlighting the need to improve readability in AI-generated medical materials for effective clinical use.