The AI loop: redefining medical research from data to clinical decisions
BMJ Digital Health & AI, 2025
Joel Grunhut
Free full text | Read at the journal | View on PubMed
In brief
An editorial on how AI is moving into every stage of medical research, from data analysis to manuscript writing to peer review. If those pieces connect, the literature could become a closed loop in which AI generates, writes and vets the evidence that guides clinical care. The piece weighs what that would mean for bias, accountability and innovation, and argues for keeping meaningful human involvement in how medical knowledge is made.
Key findings
AI first entered medical research through analysis of large imaging, health record, omics and wearable datasets, and now also drafts manuscripts and assists with peer review and editorial screening.
A closed loop, in which AI finds a result, writes it up and pre-reviews it, is not hypothetical: its components already exist in early forms.
The risks include reinforcing existing biases and health inequities, unclear accountability when AI-curated evidence misleads clinicians, and more incremental, homogeneous science.
Recommended safeguards: guidance from regulators and societies on AI in manuscript preparation and peer review, pilot programs for AI-assisted triage, explainability and auditing tools, and AI designed to surface knowledge gaps rather than reinforce known patterns.
How to cite
Grunhut J. The AI loop: redefining medical research from data to clinical decisions. BMJ Digit Health Ai. 2025;1(1):e000013. doi:10.1136/bmjdhai-2025-000013