Domain-Specific Cyberchondria and AI-Assisted Health Information Seeking among Pre-Clinical Medical Students in South India: A Two-Wave Longitudinal Study

Abstract

Background: Generative AI is widely used for health information seeking, but its longitudinal relationship with specific cyberchondria domains has not been examined. The objective was to examine domain-specific associations between AI-assisted health information seeking and cyberchondria among medical students, and to assess within-person change after AI adoption or discontinuation. Methods: Prospective two-wave study of 100 pre-clinical MBBS students at a single South Indian government medical college (January-April 2026; 100% retention). The Cyberchondria Severity Scale (CSS-15) was administered at both waves. Students were classified by AI-use status across waves into persistent non-users, new adopters, discontinued users and persistent users. Cohen’s d effect sizes, paired t-tests and Spearman rank correlations were applied. Results: AI use was reported by 81% at Wave 1 and 83% at Wave 2. AI users scored higher than non-users on Excessiveness (d=0.96), Distress (d=0.55) and Reassurance (d=0.53); Compulsion and Mistrust differed minimally. In a small, underpowered subgroup of new adopters (n=12), Distress increased significantly (Δ=+0.39, d=0.69, p=0.036); this within-person result is exploratory. Search frequency correlated with Excessiveness (ρ=0.47), Reassurance (ρ=0.57), Compulsion (ρ=0.35) and Distress (ρ=0.32). Pre-consultation symptom searching had a large effect on Excessiveness (d=1.04). Conclusion: AI users showed higher Excessiveness, Distress and Reassurance; only Distress rose significantly, and only within a small new-adopter subgroup, where it should be read as exploratory. Search frequency was the most consistent and most modifiable correlate.