Medical students who place greater trust in artificial intelligence may be significantly less likely to detect errors in AI-generated anatomical texts, according to a cross-sectional observational study published in BMC Medical Education. Researchers investigated how general trust in automated tools, analytical thinking skills, and verification habits relate to a student’s ability to identify incorrect anatomical statements. The study observed a strong negative correlation between general trust in artificial intelligence and overall anatomical classification accuracy, highlighting potential risks when automated systems are integrated into medical curricula without rigorous student evaluation.
The integration of generative artificial intelligence systems into medical education carries an inherent risk that persuasive yet inaccurate statements will be accepted without adequate scrutiny. To explore this dynamic, the study examined the relationships connecting students’ general trust in artificial intelligence, their subjective appraisal of output reliability, analytical thinking performance, self-reported verification behaviors, and classification accuracy.
Participants were presented with a 10-statement passage in which researchers had deliberately embedded factual anatomical errors in 40 percent of the items. Crucially, the students were given no prior warning regarding the presence of these inaccuracies and were asked to classify each individual statement as either true or false.
Analytical thinking tendencies were measured using the Cognitive Reflection Test, while general attitudes toward automated systems were quantified using an eight-item AI Reliability Perception Scale evaluated for internal consistency. In addition, students’ tendencies to question and verify information were recorded using a three-item rubric grounded in the lateral reading approach, and classification performance was benchmarked against an expert-determined answer key.
Multivariable linear regression and correlation analyses revealed notable divergences in student performance based on their attitudes and cognitive metrics. General trust in artificial intelligence showed a substantial negative correlation with anatomical classification accuracy (r = -0.840, p < 0.001). Furthermore, students who subjectively characterized AI-generated outputs as generally reliable performed significantly worse during the evaluation task (β = -0.924, p = 0.049), demonstrating that uncritical confidence in automated tools coincided with poorer error identification.
In contrast, analytical thinking skills and prior academic achievement emerged as independent positive predictors of accurate error detection. Higher scores on the Cognitive Reflection Test were strongly associated with improved classification performance (β = 0.957, p < 0.001). Academic standing also corresponded with greater accuracy, as students with a grade point average between 3.00 and 3.49 (β = 1.759, p < 0.001) and those with a grade point average between 3.50 and 4.00 (β = 2.300, p < 0.001) achieved significantly higher accuracy than their peers.
Despite these clear statistical associations, the authors emphasized the inherent constraints of their cross-sectional observational methodology, which cannot confirm causal mechanisms or explain the precise cognitive pathways involved. The researchers noted that while cognitive reflection, verification tendencies, and reliability perceptions are linked to error detection, further longitudinal and experimental investigations will be necessary to establish whether targeted educational interventions in analytical reasoning and lateral reading can directly enhance how students evaluate AI-generated educational materials.
Subject of Research: Evaluation of AI-generated anatomical error detection among medical students
Article Title: Medical students’ detection of errors in AI-generated anatomical content: associations with cognitive reflection and verification behavior
Article References: Taşkın Şenol, G., Büyükuysal, E. Z., & Kalaycıoğlu, O. (2026). Medical students’ detection of errors in AI-generated anatomical content: associations with cognitive reflection and verification behavior. BMC Medical Education. https://doi.org/10.1186/s12909-026-10808-4
Image Credits: AI Generated
DOI: 10.1186/s12909-026-10808-4
Keywords: medical education, artificial intelligence, anatomical classification, error detection, cognitive reflection test, analytical thinking, AI reliability perception scale, lateral reading, information verification, medical students, cross-sectional study, multivariable linear regression, academic performance, grade point average, generative AI, anatomy education
News Source: Courtney Benton. (November 16, 2026). Medical Students With High AI Trust More Often Miss Anatomical Errors. Scienmag.



