The integration of artificial intelligence (AI) within clinical research is expanding across multiple medical specialties. While discussions surrounding technological applications in healthcare are frequent, empirical data on how these tools are deployed remains critical. A recent cross-sectional study published in JAMA examines interventional trials to characterize planned artificial intelligence implementations. The study provides a view in the recent use of AI in clinical trials that provides a window in future uses of AI in clinical trials.
The cross-sectional study evaluated 1212 interventional trials registered on ClinicalTrials.gov in the last 5 years. Not surprisingly, there was consistent growth in the use of generative artificial intelligence, including large language models, rising from no reported trials in 2022 to 20.4% of trials starting in 2025, accounting for 13.7% of the total cohort. The AI applications span diverse clinical domains. The most frequent areas of focus were cancer (20.0%), neurological and mental health (19.6%), and cardiovascular and metabolic disease (18.6%). In terms of application purpose, screening and predictive modeling for disease risk or treatment selection accounted for the largest share at 39.8%. Visual and multimodal data modalities were the most prevalent, reflecting a reliance on imaging and complex datasets. Geographically, the United States accounted for 25.0% of the trials and China represented 12.3%, with a portion of studies conducted across multiple jurisdictions. Furthermore, 69.3% of the trials were single-site studies with a median target enrollment of 150 participants. At the time of data collection, 63.4% of the trials were recruiting or not yet recruiting, while 25.9% were completed.
Since the study was primarily based on how AI use was reported in the clinicaltrials.gov database, the data is incomplete. The keywords for the search were broad such as “machine learning” or “deep learning”, leading to limited information on actual use. Additionally, since the clinicaltrials.gov registry is infrequently updated, the descriptions reflect planned intentions rather than verified execution. The authors emphasize that establishing structured registry fields aligned with reporting standards would improve traceability, evidence synthesis, and reproducibility as AI integrates into clinical trial design.
The findings from this registry-based analysis demonstrate a measurable increase in planned AI implementations across interventional clinical trials. As computational and generative tools become more prevalent in protocol design, maintaining transparency is essential for accurate scientific evaluation. The absence of standardized reporting elements within current trial registries presents a challenge for tracking specific methodologies and verifying implementation. Despite these reporting gaps, not surprisingly, AI applications in clinical research are increasing at a fast pace and need rigorous scientific standards. Food for thought for the trial planners.
