Bioethics and artificial intelligence: digital content analysis on labiaplasty

Resumo

Digital marketing for labiaplasty medicalizes natural anatomical variations, necessitating large-scale bioethical screening tools. This study validates a natural language processing framework by evaluating 145 websites. The GPT-4 model, guided by bioethical directives, was compared against independent expert judgment (gold standard). The artificial intelligence demonstrated sensitivity up to 100% in identifying pathologization but null specificity (0.00%), revealing systematic disagreement with humans (Kappa between -0.06 and -0.80). A visual “blind spot” (sensitivity <40%) was identified due to the absence of descriptive metadata on websites. In conclusion, natural language processing is an impartial screening tool yet limited by informational gaps and risks of allocative harm. These results reinforce the need for multimodal approaches and mandatory metadata to safeguard patient autonomy and ethics in digital medical marketing.