Can artificial intelligence provide reliable patient education in minimally invasive bunion surgery? A comparative study of ChatGPT 5.2 and Gemini.

BACKGROUND

Patients use artificial intelligence-based large language models (AI-LLMs) to research minimally invasive surgery (MIS) for hallux valgus; however, their reliability remains uninvestigated.

PURPOSE: To compare the quality and readability of ChatGPT and Gemini responses regarding MIS bunion surgery.

METHODS

Ten frequently asked questions reflecting diverse clinical and procedural inquiries were submitted to ChatGPT and Gemini. Quality was assessed via DISCERN and 5-point Likert scales. Readability and actionability were evaluated using the Patient Education Materials Assessment Tool (PEMAT) and Flesch-Kincaid Reading Ease (FKRE).

RESULTS

No significant differences existed between models (p > 0.05). Both exceeded sufficiency thresholds for DISCERN (49.7; 49) and Likert (5.0; 4.9). While understandability (85.9%; 83.3%) and FKRE (32.9; 33.5) met requirements, both failed actionability (38.3%; 34%).

CONCLUSION

Both AI models offer reliable, high-quality theoretical information regarding MIS hallux valgus surgery. However, they are insufficient in providing actionable guidance and exceed ideal reading complexity for general patient populations.