Automate research workflows — find papers, extract data, summarize findings.
Research
Elicit is the AI research assistant built for the actual mechanics of literature review — the part where you have to find 40 papers, read them carefully, extract specific data points from each, compare them in a table, and write a synthesis. Founded in 2019 by ex-DeepMind and ex-Google researcher Andreas Stuhlmüller inside Ought (a research nonprofit) and later spun into an independent company, Elicit has quietly become the professional-grade tool that PhD students, systematic reviewers, and clinical researchers pay for when a chatbot answer is not enough and they need a real evidence-extraction workflow.
What Elicit actually does, cleanly stated, is turn a research question into a structured, exportable data extraction across dozens or hundreds of papers. You type a question — "what interventions have been studied for reducing hospital readmissions after congestive heart failure" — and Elicit returns a list of relevant papers with the abstract, the study design, the intervention, the population, the outcome, the effect size, and any other columns you ask for, all extracted into a table you can filter, export, and cite. It is not a general research chatbot. It is a systematic-review-style workflow accelerator, and it does that job better than any generalist tool on the market.
The one-line positioning: Elicit is the tool researchers use when the deliverable is a literature review, a systematic review, or a research protocol — not a chat answer. If your day includes phrases like "extract intervention type and effect size across the last 50 studies" or "screen these 300 abstracts against inclusion criteria," Elicit collapses the work by 5-10x. It is not the fastest tool for the "quick evidence look-up" question — Consensus wins that speed race. It is the strongest tool for the "produce the actual research artifact" workflow, and that is a bigger deal than it sounds.
Elicit's product is deliberately narrow and deep. The features that earn the subscription are structured data extraction, multi-paper synthesis, and the systematic review workflow. Everything else supports that spine.
Ask a research question, get a paper list. Elicit's search is grounded in Semantic Scholar's 200M+ paper corpus, tuned specifically for research questions rather than general web queries. Results are ranked by relevance to the specific question, not just keyword match.
Structured data extraction into a table. The signature feature. For any list of papers, add columns like "intervention," "outcome measured," "sample size," "study design," "effect size," "limitations noted by authors." Elicit reads each paper and populates the cells. What used to take an afternoon per 20 papers takes minutes.
Custom columns. Beyond preset extraction fields, you can define your own — "primary endpoint definition," "inclusion criteria for participants," "was the outcome adjudicated blindly." Elicit answers your columns against the paper text.
Screen and include/exclude workflow. For systematic reviews, Elicit supports the actual PRISMA-style flow — screening abstracts against inclusion criteria, marking include/exclude with reasons, exporting the resulting corpus for full-text review. This is the feature no general chatbot offers and every systematic reviewer needs.
Multi-paper synthesis. Ask a question across a set of selected papers and Elicit returns a synthesized answer with inline citations to each contributing paper. Good for the "what do these 15 studies collectively suggest about X" question that comes up at the end of a review phase.
Full-text upload and chat. Upload PDFs of the actual papers and Elicit answers questions about them — methods, results, limitations, comparisons. This turns Elicit into a reading assistant on top of a search engine.
Export to CSV, BibTeX, and Zotero. Every extraction, every paper list, every synthesis exports cleanly into the file formats researchers actually use. This is unglamorous and important.
Prompt versioning for reproducibility. Elicit lets you save and version the extraction prompts you use, so a systematic review team can share exactly how a column was extracted across every paper. Reproducibility is a first-class concern, not an afterthought.
Elicit is freemium and structured around research volume rather than model quality.
| Plan | Monthly | Annual (per month) | Included |
|---|---|---|---|
| Basic | $0 | — | 5,000 credits / month, limited exports |
| Plus | $12 | $10 ($120/year) | 12,000 credits, unlimited exports, PDF upload |
| Pro | $49 | $42 ($504/year) | 30,000 credits, priority speed, high-volume extraction |
| Team | Custom | Custom | Shared workspaces, admin, SSO |
| Enterprise | Custom | Custom | Institutional deployment, data privacy contracts |
Plus at $12/month (or $10 effective annual) is the entry tier for serious individual researchers — enough credits for a modest review project, PDF upload for reading, and unlimited export. It is the tier PhD students and freelance researchers typically start on.
Pro at $49/month (or $42 annual) is where working systematic reviewers, evidence synthesists, and clinical researchers land. 30,000 credits handles a real project — hundreds of papers with multi-column extraction. Priority speed matters when you are extracting across 300 papers in a session.
Team and Enterprise are quote-only and add shared workspaces, admin controls, and institutional data privacy contracts. Universities and pharmaceutical research groups have been signing these since 2023.
The honest read: Elicit is priced meaningfully higher than Consensus or SciSpace, and the pricing is honest — Elicit is doing more computationally per query, and the workflow value on a real systematic review project is worth 10x the Pro subscription. For casual "what does the evidence say" questions, Elicit is overkill. For actual research deliverables, it is a bargain.
Pros
Cons
Systematic reviewers and evidence synthesists. Elicit's screening and extraction workflow is genuinely modeled on how systematic reviews work. For anyone producing a Cochrane-style or PRISMA-compliant review, Elicit is a 3-5x accelerator.
PhD students on their comprehensive literature review. The extraction-to-table workflow is exactly what a chapter-2 lit review needs. Elicit does not write the chapter; it collapses the read-and-extract phase from months to weeks.
Clinical researchers scoping evidence for protocols. Before writing a study protocol, understanding the exact shape of prior evidence — what interventions have been tested, at what doses, in what populations, with what outcomes — is Elicit's home turf.
Grant writers and research proposal authors. The "background and significance" section of a grant application is a structured evidence summary. Elicit produces it faster and more comprehensively than manual PubMed labor.
Meta-analysts extracting effect sizes. For quantitative synthesis, the "extract N, mean, SD, effect size, and CI from each paper" workflow is exactly Elicit's Structured Extraction feature. Human verification still needed, but the first pass is minutes not days.
Consulting firms doing evidence-based work. Health-tech, biotech, and life-sciences consulting groups producing evidence-based client deliverables use Elicit for the same reason: it makes the extraction phase billable in hours instead of days.
Elicit's real competition is other academic AI tools and the manual workflow it replaces.
Consensus — the faster triage cousin. Consensus answers evidence questions in seconds; Elicit produces structured extractions in minutes. Different jobs. Most working researchers pay for both — Consensus for "does the evidence support X" and Elicit for "extract methodology and outcomes across 50 studies of X."
SciSpace — the paper-reading assistant. SciSpace is stronger for chatting with a single paper you have open. Elicit is stronger for structured work across many papers. Complementary, cheaper, often paired.
Perplexity with Academic Focus — general research assistant with a Semantic Scholar filter. Faster for the "quick question" flow, weaker for the "extract data across 50 papers" flow. If you want a general tool, Perplexity is broader; if you want a systematic review tool, Elicit is the specialist.
Covidence and Rayyan — the traditional systematic review software. Better than Elicit at the collaborative screening workflow for large teams; weaker at AI-powered extraction. Many teams use both — Elicit for the extraction, Covidence for the collaboration.
Zotero with GPT plugins — the DIY approach. Free, flexible, and requires you to build the workflow yourself. Fine for occasional researchers; painful at scale.
Sign up free at elicit.com. Basic tier gives you 5,000 credits — enough to run a genuine test on a real research question.
Run a real query. Type an actual research question from your work, not a toy prompt. See the paper list, add a couple of structured extraction columns, and note how the workflow feels compared to your usual PubMed loop.
Try the screening workflow. For any project where you would normally screen abstracts against inclusion criteria, upload the corpus and use Elicit's screening flow. This is the workflow the tool is really built for.
Upgrade to Plus if you are a solo researcher, Pro if you are running real reviews. At elicit.com/pricing, Plus at $12/month covers a modest research use case. Pro at $49/month is the tier for anyone doing multiple review projects a year — annual saves 15%.
Combine with a reading tool. Elicit extracts; SciSpace or Consensus surfaces evidence context. The full workflow is find (Consensus) → extract (Elicit) → read deeply (SciSpace or full PDF). Most working researchers use two of the three.
Is Elicit better than ChatGPT for research? For structured extraction and systematic review workflows, yes — and ChatGPT does not compete. For casual chat about a topic, ChatGPT is broader. The tools do different jobs.
How accurate is Elicit's extraction? For clearly reported outcomes (sample size, study design, primary intervention), high accuracy — 90%+ on well-structured papers. For complex or subtly reported outcomes (adjusted effect sizes with covariates, subgroup differences), verification is required. Elicit is a 5x accelerator, not a replacement for careful reading.
Which databases does Elicit search? Primarily Semantic Scholar's 200M+ paper corpus, which draws from PubMed, arXiv, publisher-direct feeds, and other open scholarly sources. Coverage is strongest in health, life sciences, psychology, and computer science.
Can I use Elicit for a published systematic review? Yes — but you must document your methodology. Elicit provides prompt versioning and audit trails specifically so extractions are reproducible. Best practice: run a human verification pass on Elicit's extractions before publication.
Does Elicit train on my queries and uploaded PDFs? Free and Plus data may be used for product improvement by default; you can opt out. Pro and Team contracts have stricter guarantees. Enterprise contracts specify no training use and bounded retention.
How is Elicit different from Consensus? Consensus is triage — "does the evidence support X" in seconds with a visual meter. Elicit is workflow — "extract these six data points across 50 papers and hand me a table." Different phases of research. Most working researchers use both.
Elicit Plus at $10/month or Elicit Pro at $42/month is the strongest structured-research subscription on the market — the only serious answer for AI-assisted systematic review and evidence extraction workflows. For PhD students, systematic reviewers, clinical researchers, and evidence-based consultants, the extraction-to-table workflow collapses the single most tedious phase of literature work — reading each paper and pulling specific data — from days into minutes. The reproducibility features and export formats matter for anyone whose deliverable is a real academic artifact rather than a chat log.
Where Elicit stops being the right answer: quick triage questions ("does the evidence support X" is Consensus's job in seconds, not Elicit's minutes), casual research chat (Perplexity or ChatGPT are broader), single-paper deep reading (SciSpace's chat-with-PDF UX is smoother), and non-academic research (any general tool covers it). Elicit is the specialist tool for a specific and important phase of research work, and it is priced honestly for the value it delivers on that job.
The honest recommendation: run the free Basic tier on a real research project — not a toy prompt. If the structured extraction workflow saves you hours on real deliverables, upgrade at elicit.com/pricing. Solo researchers should start with Plus at $10 annual and see if they hit the credit limits. Working systematic reviewers should skip to Pro at $42 — the higher credit budget pays for itself on a single project.