Elicit AI Review (2026): 138M Papers, a $49 Plan and the Recall Gap

A literature review used to mean weeks of reading abstracts. Elicit says it can do the screening and the data extraction for you. This Elicit AI review looks at what the tool does well, what it costs and where independent testers found gaps.

Interest in the tool keeps climbing, and searches for Elicit AI rose again this week. However, most write-ups repeat the company’s own accuracy numbers. Very few mention the one outside study that tested its search against real systematic reviews.

So the checks behind this Elicit AI review followed three questions. First, what is free? Second, what does Pro add for $49? Finally, can you trust it to find every paper?

So, what is Elicit AI? It is a research assistant built for scientific literature. You type a research question, and it returns relevant papers with a short summary of each one.

The company describes it as AI for scientific research. In addition, the homepage says it helps researchers be 10x more evidence-based. The search covers more than 138 million papers, according to the pricing page.

What sets it apart is the table. Instead of a chat window, the Elicit AI research assistant fills a grid, with one row per paper and one column per data point you ask for. As a result, it is often called the Elicit AI literature review tool.

Three groups get the most value, and their needs differ.

PhD students use the free plan to scope a topic and to find papers they missed. Meanwhile, evidence teams in pharma and medical devices use the paid review workflow. The company says 8 of the top 20 pharmaceutical firms are customers. Finally, policy analysts use the reports to get a cited overview fast.

However, it is a poor fit for some work. Researchers in theory-heavy or qualitative fields get less from extraction tables. Likewise, anyone who needs a writing tool should look elsewhere, because drafting is not the focus here.

Four parts of the product do most of the work. Here is what each one does and which plan unlocks it.

You ask a question in plain English, and the search returns papers ranked by relevance. Each result comes with a one-sentence summary written for your question.

On the free plan, search and summaries are unlimited. In addition, you can chat with papers that have full-text access and view the source behind each answer. That is generous compared with most rivals.

This is the core feature. You add a column, such as sample size or dosage, and the tool fills it for every paper in the list.

Pro allows 20 columns at a time, while Scale allows 30. Therefore a job that once took days of copy and paste becomes a few clicks. Still, every cell needs a spot check, because the model can misread what a number refers to.

The dedicated workflow guides you from search to screening to extraction. On Pro, it can screen up to 5,000 papers. Enterprise raises that to 40,000.

The company claims up to 80% time savings on systematic reviews. However, that figure comes from Elicit, so treat it as a claim and not as a measured result.

The Research Agent runs a multi-step question on its own and returns a written report with citations. Reports on Pro draw on up to 135 sources, and Scale raises the limit to 200.

Routines watch for new research and keep a project current. In addition, developers get API access from Pro up, and the site also lists an MCP server for automated workflows.

In daily use, Elicit is fast and focused. A scoping search that takes an afternoon by hand takes minutes, and the summaries are good enough to decide what to read.

The harder question is coverage. An independent study in Cochrane Evidence Synthesis and Methods tested the tool on four published reviews. On average, Elicit found 39.5% of the included studies. The original database searches found 94.5%.

Precision told the opposite story. About 41.8% of the papers Elicit returned were relevant, against 7.55% for the traditional searches. So it wastes less of your time, yet it misses more.

The authors, Oscar Lau and Su Golder, reached a clear conclusion in September 2025. Elicit could not replace a full database search at that point. Instead, they saw it as a useful extra for early scoping. For this Elicit AI review, that verdict still fits what the product is good at.

Elicit AI pricing has three individual tiers plus Enterprise. Every price in this Elicit AI review comes from the official pricing page, checked on October 7, 2026.

Is Elicit AI free? Yes, the Basic plan is free. It includes unlimited search, unlimited summaries, chat with full-text papers and Zotero import. However, the Research Agent and Research Reports come with limited usage.

Pro is the plan for systematic reviews. It adds standard usage of the agent and reports, the 5,000-paper screening workflow, 20 table columns, custom extraction from uploaded papers and API access. Scale then gives five times the usage, figure extraction, live collaboration and an admin panel.

One detail deserves attention. The $49 and $169 prices apply to yearly billing. The page says yearly billing saves 35% on Pro and 39% on Scale. By our arithmetic, that puts Pro near $75 when you pay month by month, so confirm the figure at checkout.

We could not verify any public Elicit AI promo code. In addition, the old affiliate page now returns an error, so ignore coupon sites that claim otherwise.

Elicit AI vs ChatGPT is the first comparison most people make. Both answer research questions, yet they work from different material.

ChatGPT is a general assistant that writes, codes and brainstorms. It can cite sources, but those sources are a mix of web pages and papers. Elicit, by contrast, answers from academic literature and ties each claim to a sentence in a paper.

So which should you pick? Choose ChatGPT for drafting and everyday tasks. Instead, choose Elicit when you need a table of findings that a supervisor or a regulator can trace.

Yes, Elicit is a legitimate company. It began as a project inside Ought, a nonprofit research lab, and became an independent public benefit corporation in 2023.

Andreas Stuhlmueller and Jungwon Byun founded it. Public funding records list a $9 million seed round in 2023 and a $22 million Series A in February 2025, co-led by Spark Capital and Footwork.

The homepage quotes 99.4% data extraction accuracy and more than 5 million researchers. Those numbers come from the company. In addition, Enterprise customers get a promise of no training on their data by default, plus SSO and SAML.

Two caveats stand. First, the independent search results above are far lower than the marketing suggests. Second, journals and universities set their own rules on AI in reviews, so check them first.

Consensus is a search engine that answers a question from published papers. It also shows how far the studies agree with each other.

It costs far less and suits quick evidence checks. Our Consensus AI review covers its free limits in detail.

Anara AI works from your own library of files. Every answer links to the exact highlighted passage in the source document.

Choose it for mixed reading, chat and writing at a lower price. In addition, the Anara AI review explains how its usage meter behaves.

Jenni AI is a writing tool, so it picks up where Elicit stops. It suggests text one sentence at a time and adds citations as you go.

Use it once the screening is done and the draft begins. Meanwhile, our Jenni AI review shows what the free plan includes.

Pros

• Free plan includes unlimited search and summaries across more than 138 million papers
• Extraction tables turn dozens of papers into one grid of comparable data points
• Dedicated systematic review workflow screens up to 5,000 papers on Pro
• Answers come with sentence-level citations and a view of the source
• API access and an MCP server allow automated research workflows
• High precision in independent testing, so fewer irrelevant papers to read

Cons

• An independent 2025 study measured average search sensitivity of 39.5%, against 94.5% for database searches
• Pro costs $588 a year, and paying monthly costs more
• Research Agent and Research Reports are limited on the free plan
• Live collaboration and figure extraction are locked to the $169 Scale plan
• Extracted values still need manual checks before they go into a paper

For scoping a topic and pulling data out of papers, Elicit is one of the strongest tools available. The free plan alone is worth an account for any graduate student.

Pro makes sense when you run structured reviews often. At $588 a year, it pays for itself if it saves even a few days of screening. On the other hand, occasional users will do fine on Basic.

The short answer from this Elicit AI review: use it to speed up a review, never to replace the search. Run your database queries as usual, then let Elicit catch what they missed and fill the table. The papers it never showed you are the ones no summary can warn you about.

Elicit covers screening and extraction, yet a full project also needs reading and writing tools. Our guide to the best AI research tools for papers, reviews and citations compares it with every serious rival, with prices for each.

When the evidence table is done and the draft begins, the best AI writing tools collection picks up from there.