A recent analysis reveals that among more than 1,300 artificial intelligence (AI) medical devices cleared by the U.S. Food and Drug Administration (FDA), only a negligible number have been rigorously tested for actual patient benefits. Despite widespread approval for clinical use, just three devices showed evidence of improving critical outcomes such as survival rates, stroke incidence, or quality of life.

The study assessed all AI-based medical devices authorized up to late 2025. Researchers found only a small fraction participated in registered clinical trials, with even fewer reporting results or publishing peer-reviewed findings. Most evaluations occurred in well-resourced healthcare systems and commonly excluded vulnerable groups, including pregnant women, the elderly over 75, and non-English speakers.

FDA clearance for medical AI generally relies on demonstrating “substantial equivalence” to previously authorized devices rather than proving effectiveness in improving health. This regulatory approach prioritizes rapid market entry over comprehensive validation of patient-centered outcomes. Financial and logistical challenges likely deter manufacturers from conducting extensive clinical testing. Such a gap in evidence raises concerns that these tools could unintentionally widen disparities in healthcare, especially in lower-resource settings.

The researchers warn that many low- and middle-income countries may effectively become markets for AI devices validated only through approval in wealthier nations, without independent testing in their populations. To address this, they propose a redesigned regulatory framework requiring phased evaluation to confirm benefits across diverse patient groups and healthcare environments.

According to the authors, current policies fall short of ensuring AI devices contribute meaningfully to patient health. The existing clearance process confirms similarity to existing products but does not guarantee improved outcomes. They advocate for regulatory reforms emphasizing real-world effectiveness to instill greater confidence in AI’s role in medicine.