Artificial or Authentic? A Comparative Study of Article Titles Generated by Humans and AI in Physical Medicine and Rehabilitation

Abstract

Abstract Background: This study examines expert preferences and perceptions regarding whether article titles in the field of physical medicine and rehabilitation are generated by humans or artificial intelligence (AI). It aims to evaluate the selected titles in terms of scientific accuracy, comprehensibility, originality, and attractiveness. Methods: Thirty-two article titles published in a Q1 journal indexed in the Web of Science database were selected. For each, an alternative AI-generated title was created using the ChatGPT-4o model. A survey was administered to 15 academic participants with an H-index ≥3. For each title pair, participants indicated their preference and the reason for their choice. The data were analyzed using SPSS v22 software. Results: A total of 480 title choices were recorded, with 58.1% favoring AI-generated titles. This difference was statistically significant (p=0.008). Overall, 93.3% of participants preferred AI titles more frequently. The most cited reason for preference was “comprehensibility” (47.3%). No statistically significant difference was found based on demographic variables. Conclusions: AI-generated titles were found to be strong in terms of clarity and attractiveness. However, “scientific accuracy” was cited more frequently in favor of original titles. This indicates that while AI is effective in generating appealing titles, human contribution remains necessary for ensuring scientific rigor. Hybrid models combining AI and human input may offer a balanced approach to scientific title generation.

Ähnliche Arbeiten