Elicit AI review: features, pricing, limits and who it's for

This review is for students, researchers and evidence teams deciding whether Elicit fits their work. It is part of our comparison of the best AI for deep research. Every fact about Elicit was checked on Elicit's own pages in November 2026 [1] to [7]; anything we could not confirm is marked.

What is Elicit AI?

Elicit is an AI research assistant that finds academic papers, extracts data from them into tables, and writes reports in which every claim cites a sentence from a source. It describes itself as "the most accurate AI product for scientific research" and says it supports "all AI-generated claims with sentence-level citations" [2].

Its corpus comes from Semantic Scholar, OpenAlex and PubMed, with ClinicalTrials.gov as an optional source you can select, and it is updated weekly [3]. Elicit gives the deduplicated total as over 138 million papers [1], [3]. Full text is read for open access papers. For subscription papers you need the Elicit browser extension plus your own or your library's access; otherwise Elicit works from the title and abstract [3].

How does Elicit research work?

Elicit's Research Reports follow what it calls a rapid-review process [4]:

1. Search over 138 million papers for relevant work.
2. Define screening criteria from your question.
3. Apply the criteria to filter the papers.
4. Define the data to extract from the included papers.
5. Gather that data and summarize it into a written report.

Reports "can take several minutes" and Elicit emails you when one is ready [4]. On Pro and higher you can edit the screening criteria, the included papers and the extraction columns, then regenerate, with earlier versions kept [4]. You can see and change the decisions behind the report.

What are Elicit's main features?

Elicit's features as listed on its own pages, November 2026:

What does Elicit do well?

Elicit is at its best in the middle of a review, when you know your question and need to turn many papers into comparable rows.

• Extraction tables with quotes. Each cell in an extraction table can show the passage it came from [5]. That makes checking fast: you click the cell, read the quote, and accept or fix it.
• A real systematic review workflow. On Pro, a review can screen up to 5,000 papers; Scale goes to 20,000 and Enterprise to 40,000 [1], [5]. Screening runs title and abstract first, with optional full-text screening, and every decision shows source quotes and can be overridden [5].
• Editable reports. Because you can change the screening criteria and columns behind a Research Report, you can make the report answer your question, not a generic version of it [4].
• Coverage for empirical biomedical and social science questions. Elicit says it has focused on empirical research, such as randomized controlled trials in social sciences and biomedicine [7]. Questions in those fields get its best results.

How does Elicit compare on the jobs researchers need?

We compared Elicit against five criteria that matter for academic work, using Elicit's own documentation. This is a documented comparison, not a timed trial.

Is Elicit free? Elicit pricing, checked November 2026

Yes, Elicit has a free plan, but systematic reviews and full table exports need a paid plan. These are the plans on Elicit's pricing page [1]:

The pricing page gives usage as "limited", "standard" or "5x", not as a number of reports a month [1]. Elicit's export article mentions an "Elicit Plus" plan for exporting Research Agent tables [6], which the pricing page does not list. Ask your library whether your university already has an Elicit licence before you pay.

Which Elicit plan fits?

• Basic: finding papers and reading summaries.
• Pro: one person running a systematic or scoping review, or building large extraction tables.
• Scale or Enterprise: a team that needs shared editing, dual screening or more papers per review. Dual review, where two people screen independently and Elicit shows conflicts with Cohen's kappa, is listed for Enterprise only [5].

Is Elicit AI accurate?

Elicit is reliable at pointing to real sources and less reliable at interpreting them. Its own limitations page says it "can still miss the nuance of a paper or misunderstand what a source refers to", and that it "summarizes the findings of a bad study just like it summarizes the findings of a good study" [7]. Its homepage cites a 99.4% data extraction accuracy figure from a case study [2]; that is a vendor figure from one project, not a general guarantee.

Use this five-minute check on any Elicit table before you rely on it:

1. Pick 5 rows at random.
2. For each, open the quote behind two cells: the sample size and the main outcome.
3. Confirm the quote is from the results, not the introduction or a cited study.
4. Check the population matches your question (adults vs children, country, setting).
5. If more than 1 in 10 cells is wrong, re-extract that column with a tighter instruction.

Elicit also recommends searching for papers on more than one side of a question to avoid confirmation bias [7]. See how to check a claim against the research.

Where does Elicit stop?

As described on its own pages in November 2026, these jobs sit outside Elicit or need a high tier:

• Writing the document. Elicit writes reports, not your thesis chapter or manuscript. There is no document editor for drafting in your own structure.
• Citation styles. Elicit does not format references in a chosen style. Its export guide suggests exporting a BIB file and building the bibliography in Zotero [6].
• Co-authoring. Shared Research Agent sessions and Projects exist [7], but live editing and real-time collaboration are listed on Scale and above [1].
• Judging study quality. You do this. Elicit's tables can include risk-of-bias columns you define, but the judgment is yours.
• PRISMA reporting. Elicit logs screening decisions "for PRISMA reconstruction" [5]; its help page does not describe drawing a PRISMA 2020 flow diagram [8].

Elicit vs Consensus vs AI for Research

The short version. For Consensus in depth, read our Consensus AI review.

Verdict: who is Elicit for?

• Choose Elicit if you are running a systematic or scoping review alone, or you need a large extraction table with quotes behind every cell. Pro is good value against the hours it saves on screening and extraction.
• Choose Consensus if you mostly ask focused yes-or-no questions and want a quick cited answer.
• Choose a writing workspace if your bottleneck is turning sources into a chapter or paper with correct citations.

Using Elicit with AI for Research

AI for Research covers the stages around Elicit, and some of the same ones at a lower entry price. Costs as of 12 November 2026:

1. Bring Elicit results in. Export your Elicit Library as RIS [6]. Upload it into a Systematic Review project in AI for Research (RIS, BibTeX, CSV and PubMed .nbib are accepted), where duplicates are removed by DOI, PMID and title.
2. Screen with a second reviewer. Invite a co-author as a project editor or use AI as the second reviewer. Conflicts are listed with Cohen's kappa and percent agreement. AI suggestions cost 1 credit per 25 records; the free plan allows 1 review of up to 50 records.
3. Draw the PRISMA 2020 diagram from live counts (SVG, PNG) at 0 credits, and draft the methods section for 3 credits.
4. Write the chapter. Open a Deep Research answer or a literature review draft in the editor, cite with @ from 250M+ works, and export to Word, PDF or LaTeX in any of 10,000+ citation styles.
5. Check before you submit. The citation checker confirms each reference exists and flags retractions (0 credits).

Honest limits: AI for Research's extraction table fills cells from abstracts (3 credits per 10 studies), while Elicit can extract from full text where it has it. AI for Research does not search Google Scholar or Scopus; Scopus and Web of Science exports come in as uploaded files. Free plan: 45 credits a month; Plus $12 a month for 750 credits; Pro $29 for 1,500. See pricing.

FAQ

Is Elicit free?

Yes, there is a free Basic plan. As checked on Elicit's pricing page in November 2026, it includes unlimited search across more than 138 million papers, unlimited summaries, chat with papers and Zotero import, with limited use of the Research Agent and Research Reports. Systematic reviews need Pro, at $49 a user a month billed annually ($588 a year), or a higher plan.

Is Elicit AI accurate?

Elicit is accurate at citing real papers, because every claim links to a sentence in a source. Its own help center says it can still miss the nuance of a paper or misread what a source refers to, and that it summarizes a weak study the same way as a strong one. Check a sample of extracted cells against the papers before you rely on a table.

Elicit vs Consensus: which should I use?

Use Elicit when your output is a table or a screened set of papers: it builds extraction tables with supporting quotes and runs a systematic review workflow on Pro. Use Consensus when you want a fast, cited answer to a focused question and a quick sense of how studies split. Consensus is cheaper to start; Elicit goes further into review work.

Can Elicit write a literature review?

Elicit can produce a Research Report, a cited brief built from a rapid-review process, and export it as PDF or Word. That is a strong starting synthesis, not a finished literature review. You still decide the scope and argument, weigh study quality, check each cited claim and write the chapter in your own structure and citation style.

Where does Elicit get its papers?

Elicit's help center says it searches over 138 million papers drawn from Semantic Scholar, OpenAlex and PubMed, with ClinicalTrials.gov as an optional corpus, updated weekly. It reads full text for open access papers, and for subscription papers through its browser extension and your library's access. Without full text, it works from the title and abstract.

Next steps

• Compare Elicit with the other options in the best AI for deep research.
• Read our SciSpace review if your day is spent inside single PDFs.
• Learn the method behind Elicit's review workflow in what is a systematic review.
• See where AI helps in review work in AI for systematic reviews.

Sources

1. Elicit. Pricing. https://elicit.com/pricing (checked November 2026)
2. Elicit. Homepage. https://elicit.com/ (checked November 2026)
3. Elicit Help Center. Elicit's source for papers. https://support.elicit.com/en/articles/14758040-elicit-s-source-for-papers (checked November 2026)
4. Elicit Help Center. Get a research report to generate in-depth answers automatically in Elicit. https://support.elicit.com/en/articles/14756862-get-a-research-report-to-generate-in-depth-answers-automatically-in-elicit (checked November 2026)
5. Elicit Help Center. Systematic Reviews in Elicit. https://support.elicit.com/en/articles/14759154-systematic-reviews-in-elicit (checked November 2026)
6. Elicit Help Center. Export your data from Elicit. https://support.elicit.com/en/articles/14758189-export-your-data-from-elicit (checked November 2026)
7. Elicit Help Center. Elicit's limitations; Routines in Elicit; Collaborate in a shared Research Agent session. https://support.elicit.com/en/articles/14757928-elicit-s-limitations, https://support.elicit.com/en/articles/17220392-routines-in-elicit, https://support.elicit.com/en/articles/16649941-collaborate-in-a-shared-research-agent-session (checked November 2026)
8. Page, M. J., et al. (2021). The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ, 372, n71. https://doi.org/10.1136/bmj.n71
9. Consensus. Homepage. https://consensus.app/ (checked November 2026)
10. Consensus Help Center. Subscription plans. https://intercom.help/consensus-d118a252369c/en/articles/10087865-subscription-plans (checked November 2026)