The full review
What is Consensus?
Think of it as an AI-powered academic search engine built around a simple constraint: every answer has to come from a real paper. Type a research question in plain language and Consensus searches peer-reviewed literature, then synthesizes what it finds into a cited response rather than a page of blue links. The vendor reports drawing on more than 250 million research papers, including licensed full text content from leading publishers.
That constraint is the product. Consensus positions itself as a search engine rather than an AI lab, states it does not use customer data to train models, and argues it cannot invent a source because each response is generated from an actual paper.
Deep Search turns days of reading into minutes
The flagship workflow is Consensus Deep Search, which the vendor says compresses days of literature review into a few minutes. Rather than firing one query, it builds a comprehensive search strategy on your behalf — expanding key terms, surfacing conflicting arguments, and walking the citation graph outward. The homepage examples give a good sense of the intended job: comparing randomized against observational evidence on intermittent fasting, or summarizing points of agreement and disagreement on long COVID mechanisms.
Deep reviews are metered monthly, and the allowance is what actually separates the plans. The Free tier includes a handful, Pro raises it, and the Deep plan is explicitly aimed at researchers and clinicians running literature reviews on a regular basis.
Reading agreement with the Consensus Meter
For yes-or-no questions, Consensus offers a useful shortcut. Ask whether antidepressants are effective for mild depression, and the Consensus Meter renders a visual read on how much of the retrieved evidence agrees or disagrees. It is a triage device more than a verdict — a quick way to gauge whether a field is settled before committing an afternoon to reading it. Researchers quoted on the site credit the tool with surfacing relevant papers alongside clear summaries and study designs.
Medical mode and plain-language filters
Two smaller features do a lot of quiet work. Medical mode narrows retrieval to what Consensus calls the highest-quality clinical sources — roughly 50,000 clinical guidelines and 8 million articles from the top 1,000 medical journals — which suits point-of-care questions like surveillance intervals after polyp removal. Filters are also written straight into the prompt: ask for systematic reviews from the past five years, and Consensus applies them without a sidebar of checkboxes.
Plans, seats, and campus access
Individual pricing runs Free, Pro, and Deep, with the differences expressed in Pro messages, Deep reviews, and included API and MCP uses rather than hard feature gates. Pro and Deep both unlock every research tool, including the citation graph; Deep mainly raises the monthly review ceiling and the API allowance.
Above that sit Teams, for organizations up to 200 seats with centralized billing, account management, and the Consensus Search API, and Enterprise, which adds greater volume discounts, research library integration, SAML-compliant SSO with Okta, Azure AD, OpenAthens and Google Workspace, and dedicated support. Students and faculty with a valid school email, plus US clinicians with an NPI number, can claim a discount on a subscription.
Who should choose Consensus
The best fit is people whose work is bounded by the peer-reviewed record: graduate students and PhD candidates building a dissertation base, clinicians checking guidance against current evidence, and biotech or healthcare teams tracking external literature. The vendor reports that more than 170 university libraries partner with Consensus and that students at over 10,000 universities already use it, so campus access may already exist before you pay for anything.
It is a weaker fit for two groups. If you want an assistant that writes the paper for you, set expectations carefully — Consensus positions itself as a tool that makes research more efficient rather than doing the work for you, though its Pro Analysis features can kick off an outline, a technical report, or a simple rapid review. And if your work leans on books, archives, patents, or grey literature, a corpus of journal articles will leave real gaps. Test the Free tier on a live question first; the Deep review allowance is the number to check against your workload.
Strengths
- Genuinely useful free tier: unlimited searches across 220M+ papers plus 3 Deep reviews/month
- Consensus Meter quantifies scientific agreement (yes/possibly/no) at a glance — no rival does this as cleanly
- Answers cite only real indexed papers, avoiding ChatGPT-style fabricated references
- Up to 40% off for students, faculty, and US clinicians; annual Pro works out to just $10/mo
Watch out for
- Pro's 15 Deep reviews/month runs dry fast in heavy review periods, and the next step up (Deep) jumps to $45-65/mo
- Monthly billing is steep: $15/mo is 50% over the annualized rate
- Mostly abstract-level analysis — paywalled full texts aren't included, and there's no deep linking into paper passages
- Weaker document management and data extraction than Elicit; users report browser lag in long sessions
Key features
| Feature | What it does |
|---|
| 220M+ paper database | Indexes Semantic Scholar, OpenAlex, PubMed and arXiv, updated weekly |
| Consensus Meter | Visualizes the yes/possibly/no split across studies for a research question |
| Deep reviews (Deep Search) | Synthesizes up to 50 papers into a structured mini literature review |
| Pro messages / Pro Analysis | Evidence-backed AI answers with direct citations to indexed papers |
| Study Snapshots | Per-paper summaries with methodology, sample size and quality indicators |
| Advanced filters | Filter by study design, sample size, methodology, population and more |
How Consensus compares
| Alternative | How it differs |
|---|
| Elicit | Stronger data extraction, tables and document workflows; better for systematic reviews, from ~$12/mo |
| Scite | Smart Citations show whether papers support or contrast a claim; ~$20/mo, pricier |
| SciSpace | Better for reading and chatting with individual PDFs; weaker at consensus-style synthesis |
Head-to-head: Consensus vs Elicit
Frequently asked questions
Is Consensus free?
Yes. The free plan includes unlimited basic searches across 220M+ papers, 15 Pro messages, 3 Deep reviews, and 10 Study Snapshots per month. Paid Pro starts at $120/year ($10/month) or $15 billed monthly, and students, faculty, and US clinicians get up to 40% off.
Consensus vs ChatGPT for research — which is better?
Consensus is better for academic research: it searches only its index of 220M+ real papers and cites them directly, so references can't be fabricated. ChatGPT is stronger for general writing and brainstorming but routinely invents citations. Many researchers draft with ChatGPT and verify claims through Consensus's Deep reviews and Consensus Meter.
How much does Consensus Pro cost?
Consensus Pro costs $15/month billed monthly or $120/year ($10/month) billed annually — about 33% cheaper. It unlocks unlimited Pro messages, unlimited Study Snapshots, and 15 Deep reviews per month. Verified students, faculty, and US clinicians with an NPI number can cut up to 40% off these prices.
What is the Consensus Meter?
The Consensus Meter is Consensus's signature feature: for yes/no research questions it classifies relevant studies as yes, possibly, or no, then displays the percentage split. Drawing on a 220M+ paper index, it shows in seconds whether science broadly agrees on a claim — useful for fact-checking before deeper reading.
Is Consensus good for literature reviews?
Yes, for scoping and early-stage reviews. Deep reviews synthesize up to 50 papers into structured summaries with citations; Pro includes 15 per month and the Deep plan ($45/month annual) includes 200. For full systematic reviews with data extraction tables, dedicated tools like Elicit complement it better.
Consensus ratings on other platforms
Independent user ratings from third-party review sites, linked here for transparency.
These are not our editorial score, are captured on the date shown, and may have changed since.