Elicit
Built for the unglamorous part of research — screening thousands of abstracts and extracting data into structured tables — with real, published data on where it works best.
What is Elicit?
Elicit spun out of Ought, a nonprofit machine-learning research lab, becoming a public benefit corporation in 2023 and going on to raise around $31 million, including a $22 million Series A in early 2025 co-led by Spark Capital and Footwork. Its focus is narrower and more specific than a general research search engine: rather than just answering questions, Elicit is built around the actual mechanics of a systematic literature review — screening thousands of abstracts, extracting specific data points like sample size or methodology into a structured table, and synthesizing findings across studies, following PRISMA-style workflows. More than 2 million researchers across academia, pharma, and policy work reportedly use it.
Elicit is also unusually transparent about its own accuracy, which is worth taking seriously in both directions. The company validated its abstract-screening feature against 994 Cochrane systematic reviews and reported 96.9% sensitivity — genuinely beating single-reviewer human performance on that benchmark. But an independent 2025 study published in Cochrane Evidence Synthesis and Methods (Lau et al.) found sensitivity dropped to 37.9% when researchers used search queries that matched real systematic review search strategies rather than the controlled benchmark conditions. That's a 59-percentage-point gap between Elicit's own reported number and independently observed field performance, and it's the single most important thing to know before relying on it for screening in a real review rather than exploratory reading.
The independent screening-accuracy figure comes from Lau et al., published in Cochrane Evidence Synthesis and Methods (2025) — a genuinely important methodological caveat, not a minor footnote. User and funding figures are Elicit's own disclosures as reported through mid-2026.
Key features
PRISMA-compliant systematic review workflow
Guides screening, extraction and synthesis stages that follow standard systematic review methodology.
Structured extraction tables
Pulls specific data points — sample sizes, methods, outcomes — out of many papers into one comparable table.
Research Agents
Automates multi-step research tasks across many papers rather than one search at a time.
Paper alerts
Monitors for new publications matching a topic you've defined, keeping a review current over time.
Chat with full-text papers
Ask follow-up questions directly against a paper's full text, not just its abstract.
Elicit API
Launched March 2026, letting organizations embed Elicit's search and report generation into their own tools.
Pricing
Free
- Unlimited search across 138M+ papers
- Unlimited summaries & full-text chat
- 2 automated reports / month, limited extraction columns
Plus
- Unlimited extraction columns
- Increased usage limits for students & individuals
- Full systematic review tools
Pro
- Research Agents & bulk extraction
- Full systematic review workflow
- Built for committed, ongoing evidence synthesis
A Team plan (around $79/month) and custom Enterprise pricing also exist. Exact figures for Plus and Pro vary somewhat across independent reviews ($10–32 and $49–65 respectively have both been reported) — confirm current pricing directly on Elicit's site. Annual billing saves roughly 35%.
Available models
Integrations & platforms
Pros, cons & best for
Pros
- Purpose-built structured extraction saves real hours on systematic reviews
- Genuinely generous free tier for exploratory research
- Unusually transparent about its own accuracy limitations
Cons
- Screening accuracy drops sharply under real-world search conditions per independent testing
- Weaker for theory-heavy humanities than empirical, PICO-style fields
- Narrow integrations — mainly Zotero, no Slack or enterprise workflow tools
Best for
- Graduate students and evidence-synthesis teams running real systematic reviews
- Empirical fields like biomedicine, psychology and public health
- Not the right fit for theory-heavy fields like philosophy
Take a look inside
Alternatives
For a faster first-pass answer rather than deep structured extraction:
Our verdict
Elicit is genuinely built for a specific, unglamorous, time-consuming job — turning thousands of abstracts and PDFs into a structured, comparable dataset — and for that job it remains one of the strongest tools available. What sets this review apart is Elicit's own transparency: the gap between its 96.9% controlled-benchmark screening accuracy and the 37.9% sensitivity an independent study found under realistic search conditions is a real and important limitation, not something to gloss over. Used as a first-pass accelerator with human verification still in the loop, rather than an unsupervised substitute for careful screening, it earns its reputation inside universities and research labs.
FAQ
Is Elicit's screening as accurate as it claims?
On Elicit's own controlled Cochrane benchmark, yes — 96.9% sensitivity. But an independent 2025 study using realistic systematic review search strategies found sensitivity closer to 37.9%, a substantial gap worth factoring into how much you rely on it unsupervised.
Is Elicit free to use?
Yes, the free tier includes unlimited search and summaries across its full paper corpus, though automated reports and extraction columns are capped until you upgrade.
Is Elicit good for humanities research?
Generally not as strong. It's built around empirical, PICO-style study designs common in biomedicine, psychology and public health, and is a weaker fit for theory-heavy fields like philosophy.
What is the Elicit API?
Launched in March 2026, it lets organizations embed Elicit's paper search and report generation directly into their own research workflows or products.
Does Elicit integrate with reference managers?
Yes, Zotero import is supported. Integration with broader workplace tools like Slack or Salesforce is not currently available.
Who owns Elicit?
Elicit is a public benefit corporation that spun out of Ought, a nonprofit machine-learning research lab, in 2023, and has since raised roughly $31 million in venture funding.