Elicit
Literature search and systematic review across 138 million papers, with screening and extraction accuracy figures the vendor actually publishes.
Overview
Built for the grinding part of research rather than the clever part: finding every relevant paper, screening abstracts against inclusion criteria and pulling structured data out of each one. The Research Agent plans the review, the systematic review workflow screens up to 5,000 papers on Pro, and extraction tables turn a stack of PDFs into something you can sort and filter. Users report time savings of up to 80% on review work, which is a vendor figure but a plausible one given what the manual version costs in hours.
The accuracy claims deserve some weight precisely because they are stated: 95% search recall measured against Cochrane reviews, 97% abstract screening accuracy, 96% data extraction accuracy. Those are benchmark conditions, not your topic. Still, publishing a number at all puts it ahead of competitors selling on screenshots. Basic is free and searches the whole index. Pro is $49 a month, Scale $169, and Enterprise pushes screening capacity to 40,000 papers.
An engineer evaluating this for technical literature should run a recall test on their own topic first. The index is strongest where open access is strongest, meaning medicine and psychology, and weakest across IEEE conference proceedings and industrial research where the interesting material sits behind a paywall or inside a company tech report. For due diligence on a model architecture or a sensor fusion approach it will reliably find the academic spine of the topic and miss some of the recent edges.
Key Features
- ✓ Search across 138 million academic papers, with full-text chat where the publisher permits it
- ✓ Systematic review workflow that screens up to 5,000 papers on Pro and 40,000 on Enterprise
- ✓ Extraction tables that pull structured fields into a grid: 20 columns on Pro, 30 on Scale, each cell traceable to its source paper
- ✓ Published accuracy figures: 95% search recall against Cochrane reviews, 97% abstract screening, 99% full-text screening, 96% data extraction
- ✓ Research Agent that plans a multi-step review and returns a structured report with citations
- ✓ Saved alerts for new work in a topic area, plus a Library that carries sources between projects
Where it holds
- • It publishes benchmark numbers against Cochrane reviews instead of asking you to trust a demo video, which almost nobody else in this category does
- • Extraction tables are the real product: pulling sample size, method and reported metric out of 200 papers into a sortable grid is a week of work compressed
- • The free tier searches the entire 138 million paper index, so you can test recall on your own topic before spending anything
- • Every cell links back to its source paper, so a wrong number is findable rather than mysterious
Where it breaks
- • The step from free to $49 a month is steep for someone who runs two literature reviews a year
- • Coverage skews biomedical and social science; robotics, control theory and much of the IEEE side are thinner, partly because so much of that work sits behind paywalls it cannot read in full
- • Preprints and conference proceedings, where most current AI work actually lands, are handled less confidently than journal articles
My Take
Coverage is the thing to check before anything else. The index runs to 138 million papers but leans biomedical and social science, so a robotics or control theory review turns up gaps a plain Semantic Scholar search would not. Where the money goes is extraction: point it at 300 abstracts, define columns for sample size, method and reported result, and get back a table where every cell links to the paper behind it. Elicit publishes 95% recall against Cochrane reviews and 97% abstract screening accuracy, which is more disclosure than the rest of the category bothers with.
Quick Info
- Pricing:
- freemium
- Openness:
- Proprietary
- Starting at:
- Basic is free with unlimited search across the full index and a small monthly report allowance. Pro is $49/month or $588/year. Scale is $169/month or $2,028/year with 5x agent usage and 30 extraction columns. Enterprise is custom and screens up to 40,000 papers.
- Added:
- Aug 2026
- Updated:
- Aug 2026
Use Cases
Judge it on your own work
The notes above say where Elicit holds and where it breaks. The fastest check is your own workload.
Visit website ↗Alternatives to Elicit
Gemini Notebook
freemiumGoogle's grounded research notebook, renamed from NotebookLM in July 2026. Upload sources, get answers with inline citations and run code against them in a per-notebook sandbox.
OpenEvidence
freeClinical answer engine for verified US clinicians. Cites licensed full text from NEJM, all eleven JAMA journals and 300+ other sources. Free at the point of use, funded by pharma ads.