Abstract
<title>Abstract</title> <p>Background AI-supported recommendations are increasingly discussed in dental education, but it remains unclear whether students maintain independent diagnostic judgment when external advice conflicts with their own decisions. This study examined diagnostic deference to AI-labelled and faculty-labelled advice among undergraduate dental students. Methods In a repeated-measures vignette study, 80 fourth- and fifth-year dental students completed 16 standardized pediatric and preventive dentistry cases, generating 1,280 trial-level observations. For each case, students gave an initial answer and confidence rating, reviewed an accurate or inaccurate externally labelled recommendation, and then gave a final answer. The primary outcome was harmful diagnostic deference: an initially correct answer becoming incorrect after inaccurate advice. The main secondary outcome was beneficial correction. Adjusted analyses used logistic regression with two-way cluster-robust standard errors clustered by participant and case. Results Overall accuracy increased from 57.9% to 66.4% after advice. Harmful diagnostic deference occurred in 112/374 eligible trials (29.9%), whereas beneficial correction occurred in 230/284 eligible trials (81.0%). Harmful deference was similar for AI-labelled and faculty-labelled advice (29.2% versus 30.5%), and adjusted models showed no clear association between advice label and either outcome. Conclusions External advice improved diagnostic accuracy when accurate, but inaccurate advice reversed nearly one-third of initially correct decisions. The findings do not indicate a uniquely harmful AI-label effect. Dental curricula should train students to verify and justify whether clinical recommendations should be accepted or challenged.</p>