Data for: Four Large Language Models in Parent-Oriented Education for Pediatric Flexible Flatfoot: Quality, Readability, Hallucination, and Reproducibility

Twenty parent-oriented questions on pediatric flexible flatfoot were submitted in triplicate to four large language models (ChatGPT, Claude, DeepSeek, Gemini) on 10 July 2026, yielding 240 verbatim responses. The workbook contains: (1) all 240 verbatim responses; (2) five readability indices (FRES, FKGL, Gunning Fog, SMOG, Coleman–Liau) per response; (3) independent per-response quality scores from two blinded pediatric orthopedic surgeons (four-point Likert, Mika grade, coverage, reference-standard concordance, and a hallucination taxonomy); and (4) pooled-material PEMAT-P, DISCERN, and QUEST scores from both raters. Model identities are unblinded here for transparency; during scoring, raters were blinded and responses were randomized.