What it is
Consensus is an AI search engine that answers questions using academic research. Ask something like 'does creatine improve memory?' and it finds relevant papers, summarises what they say, and shows a Consensus Meter indicating how many studies lean yes, no or mixed.
Deep reviews run a longer process that reads many papers and writes a structured literature review. Study Snapshots summarise a paper's population, methods and results. Consensus also runs inside ChatGPT, Claude and Microsoft 365 Copilot, and sells an API for embedding its search in other tools. It says more than 5 million researchers, students and clinicians use it.
Our review
Consensus gets a reader from a plain question to a cited answer faster than any general chatbot, and its weakness is built into the same feature that makes it quick. The Consensus Meter shows how many studies lean yes, no or mixed, but it counts papers; it does not weight them by sample size or design, so ten small trials can outvote one large one. For orienting yourself on a health, nutrition or psychology question that is useful. For a clinical decision it is a reading list, not an answer.
Pricing is simple at the bottom and awkward in the middle. According to Consensus's help centre, the free plan allows unlimited paper search and 3 Deep reviews a month; Pro costs $20 a month or $144 a year for 15; the Deep plan costs $65 a month or $540 a year for 200. Nothing sits between 15 and 200, and unused Deep reviews do not roll over, so a researcher who needs 30 a month pays for 200. Students, faculty and clinicians can get up to 40% off.
The API is now open to everyone. Each quick-search call returns the top 20 papers with citation counts, publication dates, study-design filters and relevance scores; Pro, Team and Enterprise include 500 calls a month, Deep includes 2,000, and further calls cost $0.05 each plus a credit per 100 papers. That makes it a cheap way to ground an internal assistant in published research, although Consensus does not publish the size of its index on its pricing or API pages.
The privacy policy is unusually direct: queries and content are never used to train AI models, Consensus's or a vendor's. Two details sit beside that promise. Voice dictation sends audio to OpenAI for transcription, and personal data is kept for as long as is reasonably necessary, with no fixed period. SSO and certifications such as SOC 2 are not mentioned on the public pages, which matters more to a university procurement office than to a student. Elicit is the stronger tool when the job is extracting the same fields from 50 papers into a table; Consensus leaves that work to the reader.
Where it earns its place
A hospital pharmacist answering a consultant's question between ward rounds
Check what trials say about a supplement interaction
Ask in plain language, read the meter, then open the two or three largest studies it cites before replying. Free paper search covers this; Pro at $20 a month adds 15 Deep reviews for longer questions.
A master's student starting a dissertation literature review
Map a field before reading in depth
One Deep review produces a structured overview and a reading list, and the student discount of up to 40% brings Pro under list price. The output is a map, not a review: the meter does not grade study quality.
An R&D engineer building an internal research copilot at a biotech
Ground the copilot's answers in peer-reviewed citations
Call the Consensus API from the copilot; each call returns 20 ranked papers with metadata and relevance scores. Pro includes 500 calls a month and Deep 2,000, then $0.05 a call, so 10,000 extra calls cost about $500.
Before you commit
- Setup
- Runs in the browser with nothing to install; paper search is unlimited on the free plan, so a first answer takes seconds.
- Integrations
- Runs inside ChatGPT, Claude and Microsoft 365 Copilot, according to its September 2026 blog; an API returns papers with metadata, filters and relevance scores.
- Data and privacy
- Queries and content are never used to train AI models; voice dictation audio goes to OpenAI for transcription; retention is 'as long as reasonably necessary'. SSO and SOC 2 status are not published.
- Cost at scale
- A ten-person lab on Pro pays $200 a month, or $1,440 a year on annual billing; Teams and Enterprise pricing is custom and not published.
- Limits
- Deep reviews are capped at 3, 15 or 200 a month by plan and do not roll over; API calls beyond the allowance cost $0.05 each.
What's good
- Cited answers from research papers in seconds
- Consensus Meter shows at a glance how the evidence splits
- Unlimited paper search on the free plan
- Clear no-training commitment in the privacy policy
- API at $0.05 a call beyond the plan allowance
What's not
- Meter counts studies without weighting size or design
- No plan between 15 and 200 Deep reviews a month
- Unused Deep reviews do not roll over
- Index size, SSO and SOC 2 status not published
Best for
Students, clinicians and curious readers who want a fast evidence check on a research question.
Alternatives
Consensus: common questions
- Is Consensus worth it?
- A fast, well-cited first pass at a research question that counts studies rather than weighing them, so the reading still falls to you. We rate it 4.2 out of 5.
- How much does Consensus cost?
- Free (unlimited paper search, 3 Deep reviews/month); Pro $20/month or $144/year (15 Deep reviews/month); Deep $65/month or $540/year (200 Deep reviews/month); Teams and Enterprise custom. Up to 40% off for students and clinicians.
- What are the downsides of Consensus?
- Meter counts studies without weighting size or design; no plan between 15 and 200 Deep reviews a month; unused Deep reviews do not roll over; index size, SSO and SOC 2 status not published.
- Who is Consensus best for?
- Students, clinicians and curious readers who want a fast evidence check on a research question.
- How much does Consensus cost for a team or in production?
- A ten-person lab on Pro pays $200 a month, or $1,440 a year on annual billing; Teams and Enterprise pricing is custom and not published.
- Is Consensus safe for company data?
- Queries and content are never used to train AI models; voice dictation audio goes to OpenAI for transcription; retention is 'as long as reasonably necessary'. SSO and SOC 2 status are not published.
- What are the best Consensus alternatives?
- The closest alternatives we have reviewed are Elicit, Scite and Perplexity.
Answers checked Oct 22, 2026.
Make Consensus? Show this rating on your site
