How AI-assisted evidence synthesis works in practice—from reproducing published meta-analyses to running new systematic reviews.
Methodology14 May 2026·12 min read
The Three-Layer Pattern: LLMs Propose, Algorithms Decide, Humans Validate
An architectural principle for agentic AI when outputs must be defensible. The LLM is the evidence-finder. A deterministic algorithm is the decision-maker. A human is the validator.
Can AI reproduce a published meta-analysis? We tested it.
One analyst, two working days. Three of four endpoints matched within 6% of a five-author published meta-analysis of neoadjuvant immunotherapy in lung cancer. The one that didn’t told us something important.