Elicit is the closest thing academia has to a default AI research assistant, used by a reported five million-plus researchers. What separates it from the discovery tools is that it carries a review past search into screening, extraction, comparison and export — the tedious middle of evidence work that everything else leaves to you. Pricing runs on credits: a free Basic tier, Plus at around $10 to $12 a month, and Pro reported between $42 and $49 monthly, which works out near $588 a year. That Pro figure is the category's sharpest wall, and casual researchers should start lower.
Reviewed by EdGrows·Updated Aug 18, 2026Affiliate link — we may earn a commission
This analysis is based on documentation, public user reports, and vendor materials — not yet on our own hands-on testing. How we rate
Most research tools help you find papers. Elicit helps with what happens afterwards, and that's the expensive part.
Screening a set of 200 results down to the 40 that matter. Reading each one for sample size, methodology, outcome measure and effect direction. Putting those fields in a table so they can be compared. Exporting the lot.
That work takes days per review and nobody enjoys it. Elicit does a first pass — extracting structured fields across dozens of papers into a comparable table, with each extraction linked back to the source passage so verification is a click rather than a re-read.
One comparison verified against live vendor pages in July 2026 put it first among AI research tools for exactly this reason: it carries a review past search and into screening, extraction, comparison and export. That's the hard part of evidence work, and almost nothing else attempts it.
This deserves care, because published figures vary more than in most categories:
| Tier | Reported | Credits |
|---|---|---|
| Basic | Free | ~5,000 |
| Plus | ~$10–12/user/mo | ~12,000, buy-more option |
| Pro | Higher, bulk extraction | |
| Scale | ~$169/mo | Team workflows |
| Enterprise | Custom | Negotiated |
Different reputable sources put Plus at $10 and $12, and Pro at $42 and $49 — one analysis verified in July 2026 states Pro bills at $588 per user annually, which lands at the $49 end. Annual billing reportedly saves up to 35%.
The spread reflects genuine plan changes over the year rather than sloppy reporting. Treat every figure here as a starting point and check the live page.
What matters structurally: credits meter the expensive operations — extraction, automated reports, high-accuracy mode — not basic search. Which means your real cost tracks how much extraction you do, and estimating from a plan description will mislead. Run one real extraction on the free tier and measure.
Pro is the category's sharpest price point, and it splits users cleanly.
Worth it: someone running systematic reviews regularly. Each review replaces days of manual extraction, and academic or analyst time costs considerably more than $588 a year. For a research group or an evidence synthesis team, this isn't a close call.
Not worth it: a graduate student doing one review for a thesis. Paying $588 for capability used twice is bad value regardless of how good the capability is.
The sensible middle, which most solo researchers land on: Plus at $10 to $12 plus free discovery tools. Semantic Scholar or ResearchRabbit's free tier for finding papers, Consensus for checking claims, Elicit Plus for extraction when it's needed.
That stack costs less than $15 a month and covers most of what a solo researcher actually does.
Worth stating plainly because it's the thing general models can't offer.
Ask ChatGPT a research question and you'll get a fluent answer that may include citations that don't exist. Ask Elicit and you get results drawn from an actual paper corpus — reported between 125 and 200 million papers depending on source — with links back to sources.
For anything heading into a thesis, a grant application or a publication, that difference isn't a preference. A fabricated citation discovered by a reviewer is a serious problem, and no amount of prompt engineering makes a general model reliable on this.
Elicit's design reflects it: extractions link to source passages, and a high-accuracy mode exists on paid tiers for work where errors matter more than speed. The correct mental model is first pass by machine, verification by human — not because Elicit is unreliable, but because that's what defensible evidence work requires.
Strongest in well-indexed empirical fields: biomedicine, psychology, social science, anything with structured abstracts and consistent methodology reporting.
Thinner in humanities, in non-English literature, and in fields where the important work lives in books and conference proceedings rather than indexed journals.
Ten minutes on the free tier tells you which side your discipline falls on, and it's worth doing before committing to anything above Plus.
Pure discovery. ResearchRabbit and Semantic Scholar do that free.
One-off reviews. Pro's price assumes repetition.
Humanities and thinly-indexed fields. Check coverage first.
Casual topic understanding. A general model is faster and cheaper, with the citation caveat above.
Against Consensus: Consensus answers whether evidence supports a claim, with a consensus meter across papers, at roughly $12 a month annually. Elicit extracts structured data across a corpus. Question-answering versus data work — Consensus is the better starting point for most people and doesn't replace extraction.
Against ResearchRabbit: ResearchRabbit maps citation networks for discovery, now with a ~$10 RR+ tier alongside its free plan. Complementary rather than competing: find with ResearchRabbit, extract with Elicit.
Against Undermind: Undermind runs deep agentic searches on one question and returns a synthesis report. Elicit gives you structured data across many papers to analyse yourself. Report versus dataset.
Against NotebookLM: NotebookLM synthesises across a fixed pack of sources you supply. Elicit searches a corpus you don't have. Different starting points.
Against ChatGPT: ChatGPT is faster and can fabricate citations. For published work that's disqualifying; for understanding a topic it's fine.
| Plan | Reported | For |
|---|---|---|
| Basic | Free, ~5,000 credits | Evaluation and light use |
| Plus | ~$10–12/user/mo, ~12,000 credits | Solo researchers, occasional extraction |
| Pro | Systematic reviews, bulk extraction | |
| Scale | ~$169/mo | Teams |
| Enterprise | Custom | Institutions |
Checked August 2026. Reported figures vary meaningfully across sources — Plus appears at $10 and $12, Pro at $42 and $49, with one July 2026 verification putting Pro at $588 per user annually. Credits meter extraction and report generation rather than basic search, and annual billing reportedly saves up to 35%. Corpus size is reported between 125M and 200M papers depending on source. Verify on elicit.com before budgeting.
Measure credits with a real extraction run. Plan descriptions won't tell you what your work costs.
Start at Plus unless you run reviews repeatedly. Pro's value depends entirely on volume.
Pair it with free discovery. Paying Elicit rates to find papers is the most common way people overspend here.
Check your field's coverage first. Ten minutes, free tier, saves a bad subscription.
Elicit is the only tool in this category that takes a literature review past finding papers and into the work that actually consumes time — screening, structured extraction across dozens of studies, comparison and export, with every extracted field linked back to its source passage. For anyone doing systematic reviews or evidence synthesis, that's a genuine day-saver rather than a convenience, and a reported five million-plus researchers reflect a tool that earned its position.
The pricing is the barrier and it's a real one. Pro reportedly runs $42 to $49 monthly, around $588 a year, which one comparison fairly called the category's sharpest wall. Credits meter extraction rather than search, so real costs track how much data work you do, and published figures disagree across sources enough that verification isn't optional. The sensible solo path is Plus at $10 to $12 alongside free discovery tools, which covers most individual research for under $15 a month.
Two things to hold onto regardless of tier. Grounding in an actual corpus is what separates this from a general model that will happily invent a citation — decisive for anything published. And coverage is uneven, strong in indexed empirical fields and thinner in humanities and non-English work, which ten minutes on the free tier will tell you.
For systematic reviewers and research teams, recommend at Pro. For students and solo researchers, Plus plus free tools is the better-value stack by a wide margin.
Note: Elicit does not currently have an affiliate program with AIVario. We earn no commission, and this rating carries no commercial incentive.
Best for: Systematic reviews and evidence synthesis, structured data extraction across many papers, research teams with recurring review work, anyone who needs citations that exist Not ideal for: Pure paper discovery, one-off reviews at Pro pricing, humanities and thinly-indexed fields, casual topic exploration Bottom line: The only tool that handles the tedious middle of a literature review properly — priced for people who do that repeatedly, and overpriced for people who don't.
Published figures disagree more than usual, which is itself worth knowing. Sources report a free Basic tier with around 5,000 credits, Plus at $10 to $12 per user monthly with roughly 12,000 credits, and Pro at $42 to $49 — one analysis verified against live pages in July 2026 puts Pro at $588 per user annually. A Scale tier around $169 monthly and custom Enterprise sit above. Annual billing reportedly saves up to 35%. Verify directly before budgeting.
Credits meter the operations that cost real compute — data extraction across papers, automated reports, high-accuracy mode — rather than basic search. The free tier's roughly 5,000 credits covers evaluation and light work; Plus adds around 12,000 monthly with the option to buy more. Extraction across a large paper set is where they disappear, which is exactly the workflow people upgrade for. Estimating from a real extraction run beats estimating from the plan description.
It handles the part after finding papers. Screening a set down to what's relevant, extracting structured fields — sample size, methodology, outcome, effect direction — across dozens of studies into a comparable table, then exporting to CSV or BIB. ResearchRabbit maps a field and Consensus answers a claim; neither does the extraction work that consumes most of a systematic reviewer's time. That's the whole reason Elicit costs what it costs.
It's built to be checkable rather than to be trusted blindly, which is the right design. Extractions link back to the source passage, so verification is a click rather than a re-read, and there's a high-accuracy mode on paid tiers for work where errors matter more than speed. The honest expectation is that Elicit does the first pass and a human checks it — treating extracted fields as final without spot-checking is how errors enter a published review.
Reported between 125 million and 200 million papers depending on source and date, which reflects both growth and inconsistent reporting rather than a contradiction. Coverage is strongest in well-indexed empirical fields — biomedicine, psychology, social science — and thinner in humanities and non-English literature. For a discipline outside the well-indexed core, checking coverage on the free tier before committing is worth the ten minutes.
Different risk profiles. A general model will answer a research question fluently and can invent citations that look plausible; Elicit is grounded in an actual paper corpus with links back to sources. For anything going into a thesis, a grant application or a publication, that grounding is the difference between a tool you can use and one you can't. For casual understanding of a topic, a general model is faster and cheaper.
For someone running systematic reviews, easily — it replaces days of manual extraction per review, and academic time costs more than that. For a graduate student doing one review a year, it's a hard sell, and the sensible route is Plus plus free discovery tools. The tell is volume: if you're extracting across dozens of papers repeatedly, Pro pays back quickly. If you're doing it once, it won't.
Free discovery plus one paid synthesis tool. Semantic Scholar or ResearchRabbit's free tier for finding papers, Consensus for checking what the evidence says, and Elicit Plus at around $10 to $12 for the extraction work. That covers most of a solo researcher's needs for roughly the price of one lunch a month, and it avoids paying Pro rates for capability you'd use twice a year.
The best sourced-answer engine there is, and the browser is now free. Pro at $20 is the tier; Max at $200 is a different product.
The fastest way to see what the literature says about a claim — provided you read the meter as a signal, not a verdict.
Still the best free AI assistant available, with no paid tier to upsell you. API pricing got more complicated in August.