Comparative Analysis of Generative AI Language Models in Orthodontics: Evidence-Based Insights Into Perplexity, iASK,...

This study is aimed at evaluating and comparing the scientific reliability of three large language models (LLMs), Perplexity, iASK, and ChatGPT 4o mini, based on their responses to orthodontic-related queries.

The three LLMs were prompted with 10 clinical orthodontic questions, and their responses were assessed independently by two evaluators using a structured scoring system (0-10). Statistical analyses, including Pearson and Spearman correlations, Cronbach's alpha, and Wilcoxon signed-rank test, were performed to determine interevaluator reliability and model performance differences.

Perplexity achieved the highest mean score (7.2), followed by iASK (5.4) and ChatGPT 4o mini (5.2). High consistency between evaluators was observed (Cronbach^'s alpha = 0.947). A significant difference was noted between Perplexity and both ChatGPT 4o mini and iASK (p = 0.002). Pearson and Spearman correlations indicated strong agreement between evaluators (r = 0.982, ρ = 1.000).

Perplexity demonstrated superior performance in orthodontic-related queries compared to ChatGPT 4o mini and iASK. The findings highlight the importance of evaluating AI models for clinical applicability and reliability.