From Prompts to Practice: Evaluating ChatGPT, Gemini, and Grok Against Plastic Surgeons in Local Flap Decision-Making

Area of Science:

• Plastic Surgery
• Artificial Intelligence
• Medical Imaging

Background:

• Local flaps are crucial for reconstructing skin defects after cancer surgery.
• Surgeon experience influences local flap selection.
• Generative AI is a potential decision-support tool with uncertain clinical utility.

Purpose of the Study:

• To evaluate the performance of three generative AI platforms in local flap selection for oncological skin defects.
• To compare AI-generated flap recommendations against expert surgeon consensus.

Main Methods:

• Three AI platforms (ChatGPT-5, Grok, Gemini) were tested using preoperative images of skin lesions.
• AI models described lesion characteristics and recommended local flaps.
• Outputs were compared to a consensus of experienced plastic surgeons.

Main Results:

• ChatGPT-5 achieved full concordance with surgeons in flap selection and accurate lesion size description.
• Grok showed intermediate performance, better at tissue plane recognition than size.
• Gemini was inconsistent in anatomical site, tissue involvement, and flap recommendations.

Conclusions:

• Generative AI can assist reconstructive reasoning from clinical images, but reliability varies.
• ChatGPT-5 emerged as the most dependable AI for local flap planning in this study.
• Further research with standardized imaging is needed to confirm clinical applicability and safety.