GatwAI

Elicit

Elicit supports scholarly research and paper analysis. Review its research agent, pricing, privacy and independent evidence on extraction limitations.

Provider sources checked

What does Elicit do?

Elicit supports finding scientific publications, extracting study information into structured fields and producing traceable research outputs. Tables help ask the same questions across multiple papers. Research Agent expands the scope to additional source types and supplied materials; Routines can prepare recurring research work. Its distinctive value is a documented path from a question to individual evidence and comparison fields. That is more demanding than a fluent summary. Systematic reviews still require independently checked search strategies, eligibility rules and search completeness. A filled table is not yet a validated evidence synthesis.
Editorial research · 2026-11-14How we research →

Who is this tool for?

  • Researchers and analysts with clear questions and documented verification methods.
  • Teams consolidating literature updates with specialist review of outputs.

Strengths

  • Repeatable extraction questions facilitate study comparison.
  • Source links support checking individual table fields.
  • Research Agent and recurring tasks can prepare maintainable evidence collections.

Limitations and requirements

  • Metadata access does not automatically provide the full paper.
  • Extraction can lose study design, endpoint or uncertainty details.
  • Comprehensive systematic searching and specialist quality assessment remain separate responsibilities.

Who is behind the product?

Company
Elicit Research, PBC
Headquarters
Headquarters not conclusively established in the primary sources used here.
Company founded
Independent public benefit corporation since September 2023; previously an Ought project.
Product launch
Documented as a literature assistant by April 2022; no exact first launch inferred from that date.

The company's spinout is distinct from the product's earlier development.

Suggested ways to use it

Suggested workflows or verification steps. These are not hands-on tests; for unverified products they are not a recommendation to use the service.

Suggested workflow: calibrate extraction on known studies

  1. Select ten relevant papers and manually recorded reference fields.
  2. Extract sample, design, endpoints and limitations using identical questions.
  3. Check every field against text or tables, record error types, then increase scope.

Goal: A controlled extraction process, not an accuracy test performed by GatwAI.

Suggested workflow: maintain a literature assessment

  1. Document the question, databases, search date and inclusion criteria.
  2. Ask for a routine proposing new evidence and changes to an existing report.
  3. Read new publications in full and accept only specialist-verified changes.

Goal: A traceable change record rather than unchecked automatic conclusions.

Pricing, allowances and data

What should you check before choosing a plan?

Basic provides free search with limited advanced capabilities. Paid tiers differ by academic versus commercial use, billing term and workload; the Industry table lists Pro at US$49 per user per month when paid annually. Compare agent, screening and extraction limits separately. Enterprise is negotiated.

What should you check about data and permissions?

August 26, 2026 help describes uploaded PDFs as encrypted, private to the account and excluded from the general corpus. This is not a blanket training or retention guarantee for every data category. Pricing explicitly lists no training by default for Enterprise; confidential research requires checking contracts, sharing and data categories.

External reviews and available evidence

Bianchi and colleagues compared extraction from 20 randomized trials with human records in 2025. Partial agreement was frequent; they call for verification and declare no conflicts of interest.

Reported positives

  • Original study with defined variables and comparison methods.

Reported cautions

  • Small bounded sample and an older product version; not evidence for all current Research Agent capabilities.

These reports reflect the linked authors’ experiences. Test date, plan and task affect results; they do not establish a representative overall rating.

Product development and notable changes

  • Routines for recurring research

    Elicit announced daily or weekly research tasks with results available for review.

    Source for this change ↗

Which alternative fits which task?

Consensus →

Consider for a quick entry through evidence-based research questions and answers.

Frequently asked questions

Can Elicit complete a systematic review alone?

The tools assist individual stages. Completeness, eligibility decisions, bias assessment and conclusions require documented specialist review.

Do uploaded PDFs become public?

Current help says they remain private and are not added to the general corpus. Other contractual and data rules still need checking.