A CROSS-SECTIONAL study has found variability in how ChatGPT (OpenAI, San Francisco, California, USA) performs when asked to generate informed consent forms for cosmetic dermatology procedures that use lasers and energy-based devices.
Informed consent is crucial in cosmetic dermatology, but producing detailed consent forms for each type of procedure can be time-consuming. The study sought to assess whether ChatGPT could offer a potential solution to manual documentation.
ChatGPT was asked to draft informed consent forms for eight commonly used laser and energy-based cosmetic dermatology devices. This included ablative and non-ablative fractional lasers, laser hair removal, pulsed dye laser, and radiofrequency microneedling.
A standardized command prompt was used as part of the study. This asked ChatGPT to provide a description of each device and procedure, and explain the risks, benefits, and alternatives.
How Did ChatGPT Perform?
General descriptions were most complete for radiofrequency microneedling (3.0), and least complete for monopolar radiofrequency devices (1.0). Benefits were best articulated for ablative fractional laser, laser hair removal, and radiofrequency microneedling energy-based devices.
Radiofrequency microneedling consistently scored highest for post-procedure course information (2.8), while ablative fractional laser scored lowest (1.6). Expected pain was well-documented for non-ablative lasers and monopolar radiofrequency devices (3.0), but less so for ablative and Q-switched lasers (2.2). Potential complications were thoroughly discussed for ablative and non-ablative lasers (2.8),but were insufficient for monopolar radiofrequency devices (1.8).
ChatGPT Variations for Recovery Time Information
The study also found significant variations in how recovery time and restrictions were covered in ChatGPT’s informed consent forms. Laser hair removal scored highest for its description of recovery time, and radiofrequency microneedling scored highest for restrictions.
How alternative treatments were covered also varied. Expected outcomes without treatment were best described for radiofrequency microneedling (2.75), but least so for non-ablative lasers (1.75), and monopolar radiofrequency devices (0.5).
Radiofrequency microneedling achieved the highest overall average rating across all categories, while monopolar radiofrequency devices had the lowest, with six out of 10 questions scoring less than two.
Expected pain and expected restrictions were communicated well. But overall impression and general description of procedures were considered to be the least complete across ChatGPT’s informed consent forms.
In summary, the study’s authors found ChatGPT shows promise in generating informed consent forms for cosmetic dermatology procedures efficiently. But variability in completeness and occasional inaccuracies does mean dermatologist oversight remains essential.
Reference:
Kim LS et al. Comparing ChatGPT informed consent documentations for energy-based devices. lasers Surg Med. 2025;DOI: 10.1002/lsm.70080. [Epub ahead of print]
