WOFAPS 2025 8th World Congress of Pediatric Surgery

Alinka Nemes-Csertő, Tamás Bűdi, Tamás Prokopp, Zita Sükösd, Zoltán Jenővári

Semmelweis University Pediatric Center, Tűzoltó Street Department, Budapest, Hungary

Purpose: The role of AI (Artificial Intelligence) in scientific publishing is expanding, raising concerns about its appropriate application. Most journals permit the use of AI to improve writing skills unless mentioned in references. However, it is also widely used for plagiarism and fabricating science. AI-generated articles can be convincing; they often even pass rigorous peer review processes. We aimed to create a paper written by AI to test its capabilities.

Method: We generated a counterfeit scientific meta-analysis of contralateral orchidopexy in pediatric testicular torsion using the free version of ChatGPT. The initial paper contained fake references, but after more specified commands, AI found existing, relevant articles for reference. The article easily passed plagiarism checkers. AI detectors only recognized 70% of the text as AI-generated.

Results: Recognizing fake publications remain challenging. Inaccurate, non-existing references, some writing patterns and inconsistency are all signs of AI fraud. Using online detectors can also not always identify AI-generated content. Noticing AI-created science strongly relies on human insight and experience; therefore, it is essential to evaluate every publication with a critical mindset.

Conclusion: While the pressure to publish continues to grow among scientists, approximately one million new papers are published each year. Based on rough estimations, 70-90% of them are not replicable, yet only a limited number are eventually retracted. Strict regulations and more profound review processes are needed to restore trust in the scientific community.