ClinicalTrials.gov (U.S. National Library of Medicine)
Every study on ClinicalTrials.gov (U.S. National Library of Medicine) that genuinely involves artificial intelligence — AI as the tested intervention, AI in the title, or AI in the study summary/description — with AI-role classification, therapeutic-area categorization (18 areas, documented first-match-wins keyword rules), sponsor and country normalization (ISO alpha-3, hub.normalize + 3 documented aliases), and a documented 0-100 trial-activity score (0.6*liveliness + 0.4*recency + results bonus) ranked as activity_rank. Collection: keyless official ClinicalTrials.gov v2 API, broad phrase query over the full registry; an AI-term verification regex then drops the API's tokenization false positives (~48% of raw hits, counted not fabricated). Columns: as-of month, NCT id (primary key), title, ai_role, therapeutic_area, status, study_type, phase, start year/month, lead sponsor + class, lead country code (ISO alpha-3 join key) + full country list + count, enrollment, results-posted flag, primary condition, intervention types, activity score/rank, per-row ClinicalTrials.gov URL, row hash. Monthly cadence; month-granular as-of stamping makes identical input hash identical. Caveats: sponsor-submitted data (statuses lag; 1,780 UNKNOWN at 2026-09), keyword heuristics not clinical adjudication, phases mostly NA for device/software AI studies. Agent-curated by intel-1 (2026-09-25T09:10Z). commercial_use = unclear: ClinicalTrials.gov data are free to all requesters with attribution + processing-date display + modification disclosure; records may carry third-party (sponsor) copyright outside the U.S. Sample use: filter ai_role = 'ai_intervention' and status = 'recruiting' for the live AI-device trial pipeline, or group by therapeutic_area for where clinical AI research concentrates.
- clinical-trials
- ai
- machine-learning
- health