Elicit is built on top of large language models — most notably Anthropic's Claude (including Claude Opus 4.5) — and a proprietary semantic search index covering 138M+ academic papers and 545,000+ ClinicalTrials.gov studies. The platform exposes multi-step AI workflows that go beyond chat: structured data extraction tables, automated abstract and full-text screening, and PRISMA 2020-compliant systematic review pipelines.

As of March 2026, Elicit provides a public API that allows developers to programmatically search the paper index and generate research reports. The system achieves 95% search recall, 97% abstract screening accuracy, 99% full-text screening accuracy, and 96% extraction accuracy benchmarked across 994 Cochrane systematic reviews. Elicit Research Agents can autonomously run multi-step workflows for topic exploration, competitive landscapes, and research summaries.