The short answer
Elicit is an AI research assistant from Elicit, Inc. that searches more than 138 million academic papers, summarises them and extracts findings into tables with citations. Since August 2026 its main entry point is the Elicit Research Agent, which also draws on trials, patents and your uploads to write cited reports. The free Basic plan includes unlimited paper search, while Pro, Scale and Enterprise add more agent usage, systematic reviews and team features.
| Made by | Elicit, Inc., United States |
|---|---|
| Free plan | Yes, Basic: unlimited paper search and summaries, limited Research Agent and reports |
| Models | Elicit Research Agent; the underlying language models are not named |
| Best at | Finding studies, extracting data into tables, systematic reviews, cited reports |
| Shows sources | Yes, with sentence-level citations |
| Team plans | Scale and Enterprise |
| Your data | Enterprise: no training by default. API: no training unless you opt in. Other plans: not stated |
What Elicit is in late 2026
Elicit is an AI research assistant from Elicit, Inc. in the United States. It started as a way to search academic papers and pull findings into a table, and that is still its core: it searches more than 138 million papers, summarises them and links every claim back to its source.
Since August 2026 it has been built around the Elicit Research Agent. Elicit announced the agent on 4 August 2026, and on 26 August redesigned its homepage so the agent is the first thing you see. Instead of choosing a workflow up front, you ask a question. The agent finds papers, weighs the evidence and writes an answer, and you can follow up in the same conversation. Systematic Review is still available from the sidebar, and since 30 September 2026 Find Papers and Extract Data also work as skills inside the agent.
The agent reaches beyond journals. Elicit says it draws on clinical trials, patents, biology and regulatory databases, conference abstracts, web sources and documents you upload, and it can produce cited reports, tables, charts, slide decks and documents. Every claim carries a sentence-level citation, so you can check exactly where it came from.
Where Elicit is genuinely strong
Answering "what does the research actually say?"
Ask a question such as whether four-day weeks raise productivity, and Elicit returns an answer grounded in studies, not blog posts. For managers who need to back a decision with evidence, this is faster and more trustworthy than a general chatbot, because the sources are papers you can open.
Turning many papers into one table
Elicit's oldest trick is still its best. Pick a set of papers and ask for columns such as sample size, method, main result and limitations, and it fills a table across all of them. Pro extracts up to 20 columns at a time and Scale up to 30. For a literature review, this replaces days of copying numbers out of PDFs.
Systematic reviews
The Systematic Review workflow screens papers against your inclusion criteria, records exclusion reasons and, since 7 May 2026, supports the PRISMA 2020 reporting guidelines. Pro screens up to 5,000 papers per review and Enterprise up to 40,000. Researchers still need to check the screening, but the first pass is far quicker.
Keeping research current with Routines
Routines, launched on 30 September 2026, let the agent watch for new evidence on a topic, update your saved work and email you a summary for review. They replaced the older Alerts feature, which Elicit retired on 31 October 2026. Routines are included from the Pro plan up.
Working as a team
Since September 2026 the Elicit Library works as a shared research hub, so a new teammate can open a project's collection and ask the agent follow-up questions straight away. Sharing with named people needs Pro or higher, sharing with the whole team needs Scale or Enterprise, and Collaborative Sessions for live joint work are on Scale and Enterprise.
Where Elicit falls short
- It is not a general assistant. Elicit will not draft your marketing email or plan your week. Use it for evidence, then take the findings to ChatGPT or Claude for writing.
- Science first. Elicit's own examples come mostly from pharma, medicine and the life sciences. For business news, competitors or current events, a web research tool such as Perplexity fits better.
- The free plan is light on the best parts. Search, summaries and chat with papers are unlimited on Basic, but the Research Agent and reports are limited.
- Papers behind paywalls. Unlimited chat applies to papers with full-text access. For others, Elicit has less to work with, so upload the PDF if you have it.
How to get better answers from Elicit
- Ask a research question, not a keyword. "Does remote work affect promotion rates for junior staff?" gets better results than "remote work promotion".
- Name the evidence you trust. Ask for randomised controlled trials, meta-analyses or studies from the last five years, and say so in the question.
- Turn on Extended Thinking for hard questions. Paid subscribers can switch it on when evidence conflicts or the question has several parts.
- Extract, do not just read. Ask for a table with the columns you care about. It is much easier to spot weak studies side by side.
- Open the citations. Check the quoted sentence for anything you will repeat in a report or a meeting. Elicit links each claim to its source, so checking takes seconds.
- Save papers to a collection. The agent can search your Library collections, and a project can be linked to one, so the next session builds on the last one.
Elicit privacy settings and your data
Elicit is most explicit about Enterprise. The pricing page promises no training on Enterprise data by default, and Elicit says it has agreements that stop third-party model providers, such as OpenAI, from training on Enterprise data. For the API, Elicit's terms, updated on 15 July 2026, say your input data is not used to train AI models unless you separately opt in. The pricing page does not state a training policy for Basic, Pro or Scale, and we found no user-facing training switch.
If you plan to upload unpublished research, patient-related material or confidential company data, ask Elicit for its terms for your plan in writing first, or use Enterprise. Elicit's API terms also say not to submit personal data through the API.
Which Elicit plan suits whom
- Basic (free): students and professionals who need to find and read papers now and then. Unlimited search makes it a strong free option.
- Pro: researchers and analysts who run literature reviews or need regular cited reports, Routines and the API.
- Scale: research teams who work on the same evidence and need five times the usage, figure extraction and live collaboration.
- Enterprise: companies and universities that need SSO, custom data sources, very large reviews and the clearest data commitments.
Check Elicit's pricing page for current prices. To see how it compares with a web research tool, read Perplexity vs Elicit, see our pick of the best AI for research, or try the lessons at Learn Elicit.
Prompts to copy
Check what the research says
What does peer-reviewed research say about [four-day work weeks and productivity]? Prioritise meta-analyses and randomised studies from the last [five] years. Summarise the strongest findings, say where studies disagree, and flag any important limitations.
You get: A cited summary that shows both agreement and doubt
Extract results into a table
Find studies on how [onboarding checklists] affect [user retention]. Build a table with columns for study design, sample size, setting, main result and limitations, and sort it by strength of evidence.
You get: A side-by-side evidence table you can scan in minutes
Find citable sources for a claim
I want to write that [sleep loss reduces focus at work]. Find peer-reviewed sources that support or contradict this claim, quote the key sentence from each, and tell me how strong the overall evidence is.
You get: Sources you can cite, plus an honest verdict on the claim
Brief a decision for your team
We are deciding whether to [introduce a hybrid work policy]. Using research evidence, write a one-page brief with the main benefits, the main risks, what the evidence cannot tell us, and three questions we should answer internally. Cite every claim.
You get: A meeting-ready brief with citations
Set up a Routine
Create a routine that checks every week for new studies on [GLP-1 drugs and workplace absenteeism], updates my [report name] report with anything relevant, and emails me a short summary of anything that changes the conclusion.
You get: A standing watch on your topic without manual searching