Elicit

Category: Research
Type: commercial (free Basic tier; paid Plus, Pro and enterprise tiers)
Source: Vendor page
Last reviewed: 2026-10-17 (what this means)

Summary

Elicit is a commercial AI research assistant that searches academic papers and supports systematic-review steps (screening and data extraction). Its literature search is weak. In independent tests, mostly in health research, Elicit searches found about 18–40% of the studies identified by traditional systematic searches, and in one 2025 test only 8% of its results overlapped a Boolean search. Its data extraction agreed with human reviewers about 77–92% of the time in most tests but was accurate in only 51% of responses in a 2024 test, with some errors and made-up details; screening caught 71–100% of relevant studies in two recent tests. Researchers concluded that Elicit can assist but not replace human reviewers. Most tests used paid tiers between 2023 and 2026. Overall score 68. Informational only; not a guarantee, endorsement, or certification.

Overall score: 68 · Evidence grade: A · Independent sources: 10 performance, 2 other

Evidence grade describes the amount and type of evidence, not the agent's quality.

Factor breakdown

Factor breakdown
Factor	Weight	Score	Note
Completion rate	30%	55	Search: average sensitivity 39.5% vs 94.5% for traditional searches (Lau et al.), 17.6% of included studies found (Bernard et al.), 1 of 59 results relevant (Vyas et al.), 158 of 5,285 studies (Wei et al.). Screening: sensitivity 0.71 (abstract) and 0.78 (full text), precision 0.38 (Wei et al.). Extraction: 92% precision/recall (Andersen et al.), 77–90% value matches (Lagisz et al.), 79% fully accurate (Wei et al.), mostly partial matches for complex variables (Bianchi et al.). Mazzali et al. (Pro, Sept–Oct 2025): search overlapped 8% with a Boolean search; screening recall 90.9% (abstract) and 100% (full text). Meliante et al. (March 2024): search found 24.5% of included studies; extraction accurate in 51.4% of responses. Spillias et al. (January 2024): extraction quality above the acceptable threshold and best of three AI setups. Judgment call.
Disputes	20%	80	No legal or regulatory actions, security advisories or data incidents were found. Fabricated details found in tests (4 confabulations in Andersen et al.; a wrongly cited protocol in Bernard et al.) are scored under completion. Judgment call.
Adherence to limits	25%	55	Results varied between repeated runs (169–246 search results across three identical searches; extracted values matched 90% across accounts, but supporting quotes only 46%). It sometimes gave multiple answers where one was required and defaulted study location to the authors' country. Privacy and no-training commitments for uploaded files are vendor statements only. It answered nearly every extraction question even when the paper contained no relevant data (27% false positives; Spillias et al.), and the Systematic Reviews workflow was capped at 500 papers per search (Mazzali et al.). Judgment call.
Evidence quality	25%	85	Ten independent performance sources, nine peer-reviewed, one testing the current paid tier in 2025–2026. Samples are small and mostly from health and environmental sciences; two tests date from early 2024, and several do not state the tier. One had free accounts from Elicit (partly independent). Two further studies were read from abstracts only and are not counted. Judgment call.

Overall = 0.30×55 + 0.20×80 + 0.25×55 + 0.25×85 = 67.5; an exact .5 is rounded up, so it is shown as 68. See the methodology.

Scope

This entry covers Elicit, all tiers, tested mainly on paid tiers (Plus, Pro and the Review/Systematic Review workflow). The free Basic tier was not tested separately. Elicit has changed often (a move from credits to subscriptions in July 2024, high-accuracy extraction by default from May 2025, and a Systematic Review workflow); tests from 2023 to 2026 are counted because the product is continuous. The results differ by task: searching is weak, screening is moderate and data extraction is stronger, so the overall score does not describe fitness for any single task.

Privacy and disputes

No lawsuit, regulatory action, security advisory or reported data incident was found. Elicit states that uploaded PDFs stay private to the user’s account, that it does not train on enterprise data, and that it holds a SOC 2 Type II report. These are vendor statements only; no independent public audit was found.

Uncertainties

Most tests used paid tiers (Plus, Pro, Review/Systematic Review workflow); the free Basic tier was not tested separately.
Elicit changed often (credits to subscription in July 2024; high-accuracy default from May 2025; Systematic Review workflow); the oldest study used April 2023 searches.
Two studies (Hilkenmeier et al., O'Rourke et al.) are listed but not counted: abstract only, not fully reviewed (full text paywalled).
Lagisz et al. used temporary free accounts provided by Elicit (no other Elicit involvement, per the authors); counted and labelled partly independent. Spillias et al. include a UNSW author; there is no author overlap with Lagisz et al.
Studies are small and mostly in health sciences; results differ by task (search, screening, extraction).
The vendor's own evaluation (93.6% recall, 94–99% extraction accuracy) is not used.
The overall score is exactly 67.5; exact .5 rounds up (Decision 0010).
Grade A rests on one recent test (Wei et al., preprint) that states the Pro tier and the dates of use; Elicit publishes no version numbers.
Extraction results vary widely between studies (51–92%); the lowest result comes from a March 2024 test, before the high-accuracy default.

Sources

Excluded sources

Elicit’s own evaluation (93.6% screening recall, 94–99% extraction accuracy): vendor-run, not independent.

Decision records: 0051 (grade-A basis check); 0054 (citation backfill).