Elicit
Elicit is a systematic review workflow tool with a strong structured extraction interface, but independent evaluations find its search sensitivity averages 39.5% and its quote/reasoning matching degrades to 10%/0% in high-accuracy mode, positioning it as a complement to human reviewers rather than a replacement.
What is Elicit?
Elicit is an AI research assistant designed around systematic literature review rather than general-purpose research. Its core workflow takes a research question, searches academic databases, and structures the results into a table with extracted data fields—study design, sample size, intervention, outcome—so that papers can be screened and compared without reading each one in full. The platform supports Research Reports, a Research Agent, and a dedicated Systematic Review workflow that the Pro plan markets as capable of screening 5,000 papers.
Elicit’s positioning is narrower than general AI research tools. It is not a citation-context analyzer like Scite, nor a claim-agreement checker like Consensus. It is built for the process of conducting a review: identifying candidate papers, extracting their characteristics, and assembling the extracted data into a usable evidence table. The pricing reflects this orientation: the free Basic tier provides unlimited search and summaries but caps automated reports at 2 per month and provides only a one-time 5,000-credit allocation for advanced features. The workflow that Elicit is actually designed for—screening thousands of papers and extracting 20+ columns of data—requires Pro or Scale.
The independent evidence on Elicit’s performance is more sobering than its product marketing. A 2025 peer-reviewed comparison across four systematic reviews found Elicit Pro’s sensitivity averaged 39.5%, compared with 94.5% in the original human-conducted searches[reference:19]. Elicit did identify some studies the original searches missed, and its precision was substantially higher (41.8% vs. 7.55%), but it cannot be relied on as a primary search method. A 2026 feasibility study on data extraction found that value matches could exceed 87% on tuned prompts but that quote and reasoning matches were far lower, and dropped to 10% and 0% respectively in high-accuracy mode[reference:20]. Elicit also cannot extract data from figures, which excludes a significant portion of quantitative results in medical and life-science reviews[reference:21].
How does Elicit work?
Users begin with a research question entered in natural language. Elicit searches its indexed corpus—138 million papers and 545,000 clinical trials on paid plans—and returns a list of candidate papers with short AI-generated summaries. From there, the user can add extraction columns, defining the data fields they want populated for each paper. Elicit attempts to extract those fields automatically, displaying the value alongside the supporting quote and source paper so the user can verify the extraction.
The Systematic Review workflow on Pro adds screening capabilities: inclusion and exclusion criteria can be defined, and Elicit will screen up to 5,000 papers against them. Research Reports generate a narrative summary with citations, and the Research Agent performs a broader multi-step investigation. Scale adds figure extraction, real-time collaboration, and the ability to extract from up to 200 data sources. Enterprise adds PRISMA-grade screening, 40,000-paper screening capacity, SSO/SAML, and no training on user data by default.
API access is available starting at Pro. The API draws from the same credit pool as the platform plan and permits building “Downstream Products” that incorporate API outputs, subject to restrictions against reselling standalone API access or building a competing search engine[reference:22]. It does not permit proxying API requests or sharing keys with third parties outside the organization[reference:23].
Use Cases
- Conducting preliminary scoping searches where Elicit’s higher precision (41.8%) is useful for identifying relevant papers quickly, with the understanding that it will miss a majority of eligible studies and cannot substitute for a comprehensive database search
- Generating a first-pass extraction table for a systematic review, then having a human reviewer verify every value and quote before inclusion
- Screening a large batch of papers against inclusion criteria in the Pro Systematic Review workflow, with human adjudication of borderline cases
- Creating structured evidence tables with 20–30 columns for comparison across studies on Scale
- Running Research Reports for narrative summaries that cite their sources, subject to the 2-report Basic cap or higher paid limits
Pricing
Pricing model: Freemium
| Plan | Price | Billing | Included usage | Notes |
|---|---|---|---|---|
| Basic | $0 | free | Unlimited search across 138M+ papers, unlimited summaries, unlimited chat with papers with full-text access, view sources, Zotero import. Limited usage for Research Agent and Research Reports. A one-time allocation of 5,000 credits is included; it does not refresh monthly. Basic caps automated reports at 2 per month and extraction columns at 2 per month. | — |
| Plus | $132 | per year | Per user/month, billed as $132 annually. Everything in Basic plus export to RIS, CSV, BIB, PDF, and DOCX; add 5 columns to tables at a time; unlimited search across 500,000+ clinical trials; standard usage for Research Agent, Research Reports, and Systematic Literature Reviews. | — |
| Pro | $468 | per year | Per user/month, billed as $468 annually ($49/month effective). 5× usage for Research Agent, Research Reports, and Systematic Literature Reviews. Dedicated Systematic Review Workflow that can screen 5,000 papers. Add 20 columns to tables at a time. Reports extract from up to 135 data sources. Access to Routines. Custom extractions from uploaded papers. Explanations for AI-generated answers. Multiple output templates. API access. | — |
| Scale | $1,068 | per year | Per user/month, billed as $1,068 annually ($89/month effective; some sources report $169/month list price). 9× usage for Research Agent, Research Reports, and Systematic Literature Reviews. Extract and interpret from figures in research papers. Live editing and real-time collaboration. Reports extract from up to 200 data sources. Add 30 columns to tables at a time. Admin panel with usage tracking and seat management. | — |
| Enterprise | Custom | custom | Custom usage for Research Agent, Research Reports, and Systematic Literature Reviews. No training on your data by default. Screen 40,000 papers and extract 40 columns. PRISMA-grade screening and extraction accuracy. Enterprise security including SSO/SAML, 2FA, user analytics, domain verification, single-tenancy deployments. Dedicated customer success. Unlimited API access. | — |
Common Limitations
- Search sensitivity is poor in independent evaluations, limiting its value as a primary search tool · A 2025 peer-reviewed comparison of Elicit Pro against traditional systematic review searches across four case studies found that Elicit’s sensitivity averaged 39.5% (range 25.5–69.2%), compared with 94.5% (91.1–98.0%) in the original human-conducted reviews[reference:0]. Elicit’s precision was higher (41.8% vs. 7.55%), meaning the papers it does return are relevant, but it misses a majority of the studies that a thorough search would identify. The authors concluded that Elicit “did not search with high enough sensitivity to replace traditional literature searching,” though it identified some studies the original searches missed and could serve as a useful adjunct for preliminary searches[reference:1]. A tool that returns relevant papers but misses more than half of the eligible evidence cannot be used as the sole search method for a systematic review.
- Data extraction accuracy is moderate at best, and reasoning/quote matching degrades sharply · A 2026 peer-reviewed feasibility study across environmental and life sciences found that of 90 considered extraction prompts, 70 exceeded 87% accuracy when compared to gold-standard human extraction, but accuracy tended to be lower when tested on a new set of articles[reference:2]. Repeating extractions with different Elicit accounts produced 90% agreement on extracted values, but supporting quotes and reasoning matched in only 46% and 30% of cases respectively[reference:3]. In high-accuracy mode, value matches dropped to 77%, quote matches to 10%, and reasoning matches to 0%[reference:4]. The study concluded that Elicit “can complement, but not replace, human data extractors” and may be best used for sanity checks and protocol clarity evaluation rather than as an authoritative extraction source[reference:5]. A separate evaluation found Elicit’s extractions for nuanced variables such as intervention effects were “less detailed, with 95% rated as partially equal”[reference:6].
- Unable to extract information from figures and flowcharts · Elicit cannot extract data from figures, including flowcharts and graphical presentations of results[reference:7]. For systematic reviews in medicine and life sciences, where participant flow diagrams (CONSORT, PRISMA), Kaplan-Meier curves, and forest plots carry critical numerical data, this is a structural limitation rather than an edge case. The Scale plan adds “extract and interpret from figures” on the official pricing page, but independent evaluations of the Pro tier note figure extraction as a current limitation, and the capability is plan-gated rather than available across tiers[reference:8].
- Basic plan credits are one-time and do not refresh, making the free tier a trial rather than a sustainable allowance · The Basic plan includes 5,000 credits as a one-time allocation, not a monthly refresh[reference:9]. Third-party cost analyses note this contrasts with competitors whose free tiers refresh monthly, and that once the initial credits are exhausted, advanced features require a paid upgrade[reference:10]. Automated reports are separately capped at 2 per month on Basic, and extraction columns at 2 per month. The free tier is sufficient to evaluate the interface and search behavior but not to conduct even a small-scale review without purchasing a paid plan.
- Commercial use is permitted but the license scope is “internal business purposes,” and data-training status is not disclosed · Elicit’s Terms of Service grant a license to use the Platform “solely for User’s internal business purposes”[reference:11]. This means commercial use within an organization is permitted, but the license is not framed as a general commercial grant for client-facing products or resale of outputs. The API terms separately permit building “Downstream Products” that incorporate API Outputs, subject to restrictions against reselling standalone API access or building a competing search engine[reference:12]. On data training, Elicit’s status is listed as “not disclosed” in third-party comparisons, which is a silence rather than a commitment; Enterprise includes “No training on your data by default,” but lower tiers do not carry that explicit default[reference:13].
- Pricing is split between academic and industry rates, and the difference is large · Elicit maintains separate price books for academic and industry users. Third-party pricing analyses report that the same Scale feature set costs approximately $49/month for academics versus $169/month for industry, and that academic pricing runs “far lower” across tiers[reference:14]. The public pricing page shows annual billing figures without clearly labeling which price book applies, and monthly-versus-annual report allocations differ: Plus annual provides 48 reports/year (4/month average) usable unevenly, while monthly billing strictly limits usage to 4/month, reducing flexibility at the same nominal cost[reference:15]. Users should confirm which price book their account falls under before budgeting.
Key Features
- Search across 138M+ papers and 545,000+ clinical trials
- Structured data extraction into tables with supporting quotes and source links
- Systematic Review Workflow on Pro, screening up to 5,000 papers
- PRISMA-grade screening on Enterprise, up to 40,000 papers
- Research Reports and Research Agent with usage tiered by plan
Elicit vs. Consensus and Scite
This comparison focuses on documented differences in primary function, pricing orientation, and the specific evidence each tool produces. It does not rank overall research value, which depends on the review stage and the question being asked. Elicit, Consensus, and Scite address different parts of the literature workflow; they are not interchangeable substitutes.
| Tool | Positioning | Key differences | Best fit | Pricing note |
|---|---|---|---|---|
| Elicit Pricing Search sensitivity evaluation (Wiley, 2025) Data extraction feasibility study (2026) | A systematic review workflow tool built around paper search, structured data extraction, and a dedicated screening interface with PRISMA support on higher tiers. | Optimized for the extraction-and-screening stages of a systematic review rather than for rapid claim verification or citation-context analysis. Independent evaluations show poor search sensitivity (39.5% average) but higher precision than traditional searches. Extraction value accuracy can exceed 87% on tuned prompts, but quote and reasoning matching degrades sharply in high-accuracy mode, and figure extraction is plan-gated. Free tier provides unlimited search but only one-time credits and 2 automated reports per month. | Researchers who need a structured extraction table and screening interface for a review they are already conducting through conventional database searches, and who will verify every extracted value against the source. | Basic free (unlimited search, 5,000 one-time credits, 2 reports/month). Plus $132/year. Pro $468/year ($49/mo effective), Systematic Review Workflow screening 5,000 papers, API access. Scale $1,068/year ($89/mo effective, some sources list $169/mo), figure extraction, collaboration. Enterprise custom. |
| Consensus Comparison context (The AI Agent Index, 2026) University of Maryland comparison guide | An academic search engine oriented toward answering specific research questions by showing where peer-reviewed studies agree or disagree, with a Consensus Meter. | Consensus is built for rapid claim verification and evidence direction (“does the literature support X?”), not for structured data extraction or systematic review screening. Its distinguishing feature is the Consensus Meter, which aggregates study findings into an agreement/disagreement summary. Pricing is substantially lower: Pro starts at approximately $12/month billed annually, with a free tier that is more usable for ongoing work than Elicit’s one-time-credit Basic plan. Consensus does not offer Elicit’s systematic review workflow or extraction-table interface. | Clinicians, students, and researchers who need a fast read on what the peer-reviewed literature says about a specific claim, rather than a structured extraction and screening workflow. | Free tier available. Pro approximately $12/mo billed annually. Higher Deep plan available. MCP server included. |
| Scite Cardiology comparative evaluation (2026) University of Maryland comparison guide | A citation-context analysis tool that classifies how a paper has been cited—supporting, contrasting, or mentioning—rather than searching for or extracting from papers. | Scite’s unit of analysis is the citation, not the paper. It reports whether subsequent literature supports or contradicts a claim in a source, and flags retractions. It does not provide Elicit’s structured extraction tables or Consensus’s agreement meter. A comparative evaluation across cardiology topics found Scite performed lowest among tools tested on proportion of relevant articles identified[reference:33], but that evaluation measured search relevance, not citation-context accuracy, which is Scite’s specific strength. Pricing starts at approximately $15/month for individual access. | Researchers auditing how a specific paper or claim has been treated in the subsequent literature, including whether it has been supported, disputed, or retracted. | Free basic plan. Premium from approximately $15/month. Institutional licenses available. |
Comparison checked against official product information on 2026-11-05. Features and pricing may change.
Frequently Asked Questions
Is Elicit accurate enough to replace manual systematic review searches?
No. A 2025 peer-reviewed comparison across four systematic reviews found Elicit Pro’s sensitivity averaged 39.5%, compared with 94.5% in the original human-conducted searches[reference:24]. Elicit returned relevant papers at a higher precision rate (41.8% vs. 7.55%) and identified some studies the original searches missed, but it missed a majority of eligible studies. The authors concluded that Elicit’s searches were “not sensitive enough to replace traditional searching” and recommended it as an adjunct for preliminary searches[reference:25].
How accurate is Elicit’s data extraction?
A 2026 feasibility study found that 70 of 90 considered extraction prompts exceeded 87% accuracy against gold-standard human extraction when tested on the same articles, but accuracy tended to be lower on a new set of articles[reference:26]. Repeating extractions with different accounts produced 90% agreement on values, but supporting quotes and reasoning matched in only 46% and 30% of cases. In high-accuracy mode, value matches dropped to 77%, quote matches to 10%, and reasoning matches to 0%[reference:27]. The study concluded that Elicit can complement, but not replace, human data extractors.
Can Elicit extract data from figures and charts?
On the Pro tier, independent evaluations report that Elicit cannot extract information from figures, including flowcharts and graphical results[reference:28]. The Scale and Enterprise plans list “extract and interpret from figures” as a feature on the official pricing page, but figure extraction is plan-gated rather than available across tiers. Users conducting reviews where data is presented primarily in figures should confirm figure-extraction availability on their specific plan.
Does Elicit’s free plan have monthly credits?
No. The Basic plan includes 5,000 credits as a one-time allocation, not a monthly refresh[reference:29]. Automated reports are separately capped at 2 per month, and extraction columns at 2 per month. Once the 5,000 credits are exhausted, advanced features require a paid plan. The free tier is useful for evaluating the interface and search behavior but not for conducting a full review.
Can I use Elicit for commercial work?
Elicit’s Terms of Service grant a license to use the Platform “solely for User’s internal business purposes,” which permits commercial use within an organization but does not frame the license as a general grant for client-facing products or reselling outputs[reference:30]. The API terms separately permit building Downstream Products that incorporate API outputs, subject to restrictions against reselling standalone API access or building a competing search engine[reference:31]. For client deliverables or productized outputs, the internal-business-purposes framing should be reviewed against the specific use case.
Alternative Tools
-
NotebookLM NotebookLM (now Gemini Notebook) is Google's source-grounded research assistant: you upload documents, and it answers only from those sources with inline citations, then turns them into Audio Overviews, Video Overviews, Mind Maps, Slide Decks, and study tools. The free tier is genuinely usable; paid limits are bundled inside Google AI plans. The hard boundaries are no consumer API, answers bounded by uploaded sources, and Audio Overview language quality that is still catching up outside English. -
Exa Exa is a semantic search API built on its own 500B+ page index, with page text included in search results, strong entity-lookup accuracy, an official MCP server, and per-request pricing from $4/1k. It is not a benchmark leader across the board—it is statistically tied with Brave and Firecrawl on independent agentic search benchmarks—and its 10 QPS limit, non-paginating search, crawl-bounded recency, and Enterprise-only compliance features are hard constraints that require workflow design around them. -
Perplexity Perplexity is a cited answer engine that combines real-time web search with advanced language models, but its $20 Pro tier is licensed for personal non-commercial use only, its Pro Search allowance is an unpublished “weekly limit,” and its API is a separate pay-as-you-go billing system. -
Tavily Tavily is a web access API for AI agents with a genuinely usable free tier (1,000 credits/month, no card), bundled full-page text in search results, and an official MCP server. But its Research endpoint is dynamically priced at 4–250 credits per request, the headline 1,000 RPM applies only to paid Production keys, entity and academic retrieval lag specialized indexes by large margins, and query data may be used for training unless Zero Data Retention or the data-use setting is changed. -
Brave Search API Brave Search API provides web, news, image, video, local, and LLM Context results from Brave's independent index, priced per request with a $5 monthly credit, aimed at AI agents and RAG pipelines rather than Google SERP replication. -
Adobe Firefly Adobe Firefly is a generative AI platform built around commercially safe, licensed training data and deep integration with Adobe’s creative applications, offering image, video, audio, and partner-model generation through a credit-based subscription system.