Consensus: What It Is, Pricing and How to Use It (2026 Guide)

What is Consensus?

Consensus is an AI search engine that only searches peer-reviewed scientific literature. You type a research question in plain English, and instead of ten blue links from the open web, you get a list of relevant papers, an AI-written answer that cites those papers, and tools for digging into each study. The vendor describes it as “an AI academic search engine for peer-reviewed literature” and says it searches more than 400 million papers.

The product is made by Consensus, a US company, and runs in the browser. Its pitch is that general AI assistants will answer any question confidently whether or not the evidence supports it, while a search over the scholarly record forces every claim back to a source you can open and check. The company says more than 5 million researchers, students and clinicians use it.

The problem it solves is the gap between “I want to know what the research says” and the hours it takes to find, screen and read papers in Google Scholar or PubMed. Consensus does the finding and the first pass of screening for you, so you spend your time on the papers that matter.

Who it is for

• Students and academics who need a fast first pass over the literature before a deeper review in a library database.
• Clinicians, nurses and allied health professionals checking whether an intervention has evidence behind it.
• Analysts, consultants and journalists who want citable studies rather than vendor white papers.
• Content writers and marketers who make health, nutrition, psychology or technology claims and need a source for each one.
• Product and UX teams looking for research on behaviour, design or productivity questions.

Consensus is not the right tool for a formal systematic review, where you need reproducible search strings, screening logs and appraisal of every included study. It is also the wrong place for questions the research does not cover, such as pricing of a product, current news or company-specific data. For those, a general search tool is better. See our comparison of Perplexity vs ChatGPT for research for the web-search side of the picture.

Key features

• Pro search. The default mode. Ask a question and Consensus returns relevant papers and a cited AI analysis that synthesises what they found. On the pricing page these are counted as “Pro messages”.
• Deep search. A longer, literature-review style mode that works through many more studies and produces a structured report with findings and gaps. The pricing page counts these as “Deep reviews” and caps them per month on every plan.
• Consensus Meter. For yes/no questions, a bar that shows how the top papers lean (for example yes, possibly, mixed or no). It is a quick orientation on where the evidence sits, built from a small set of top results, not a vote across the whole field.
• Study Snapshots. A summary card for each paper that pulls out key findings and methodology details such as population, sample size and duration, so you can decide whether a paper deserves a full read.
• Ask Paper. Chat with a single paper. Ask about its methods, limitations or whether a result applies to a specific population, and the answers come from that paper’s text.
• Citation Graph. Shows relationships between papers, which is useful for finding the influential studies a result builds on or the later work that cites it.
• Filters. Narrow results by criteria such as publication date and study characteristics so the analysis is based on the studies you care about.
• Collections and citation export. Save papers into Collections to organise a project, and export citations in RIS, BibTeX or CSV for your reference manager.
• API and MCP access. Paid plans include a monthly allowance of API and MCP uses, so you can call Consensus from scripts or from AI assistants that support MCP.

Pricing

At the time of writing Consensus has a Free plan and two paid individual plans, plus Team and Enterprise options for universities and organisations on request. The prices below are what the pricing page shows with annual billing; the page states that annual billing on Pro saves $96 a year against monthly billing, so expect the monthly rate to be higher. Check the pricing page toggle before buying.

Plan	Price	What you get	Who it suits
Free	$0	Basic paper search, 10 Pro messages a month, up to 3 Deep reviews a month	Trying it, occasional questions
Pro	$12/month billed annually ($144/year)	Unlimited Pro messages, 15 Deep reviews a month, access to every research tool, 500 API and MCP uses a month ($0.05 per use after, up to 1,000)	Students, writers and professionals using it weekly
Deep	$45/month billed annually ($540/year)	Everything in Pro, 200 Deep reviews a month, 2,000 API and MCP uses a month ($0.05 per use after, up to 10,000)	Researchers and clinicians running frequent literature reviews

Students and faculty with a valid school email, and US healthcare professionals with a valid NPI number, can claim up to 40% off. Many university libraries also provide institutional access, so check with your library before paying.

Deep reviews are the number to watch. They are the most useful feature for serious work and the most tightly rationed, so save them for well-formed questions.

How to use Consensus: step-by-step

1. Create an account and run a first search

1. Go to consensus.app and create a free account.
2. Type a question into the search bar. Write it as a real research question rather than keywords: “Does intermittent fasting improve insulin sensitivity in adults with type 2 diabetes?” works better than “intermittent fasting diabetes”.
3. Leave the default Pro mode selected and run the search.
4. Read the AI analysis at the top. Each sentence is tied to one or more papers; click the citations to open the source.
5. Scroll to the result list. For a yes/no question you will also see the Consensus Meter above the results.

2. Triage the results

Most of the value of Consensus comes from what you do after the first answer.

• Open the Filter panel and narrow by publication date, study characteristics and other criteria. Rerun the analysis so it reflects the filtered set.
• Expand a Study Snapshot on any result to see population, sample size, duration and the main finding. This takes seconds and tells you whether the paper is a randomised trial of 2,000 people or a survey of 40 students.
• Use Ask Paper on the two or three most relevant studies. Good questions: “What were the limitations the authors reported?”, “Was the effect still significant after adjusting for confounders?”, “What population was this tested on?”
• Open the Citation Graph on a key paper to find the studies it builds on and the later work that cites it, which often surfaces the most recent evidence.

3. Run a Deep review for a proper literature pass

When a question matters enough to justify one of your monthly Deep reviews:

1. Write the question carefully first. Include the population, the intervention or variable, and the outcome. Vague questions produce vague reports.
2. Switch the search mode to Deep and run it. It takes noticeably longer than a Pro search because it works through many more studies.
3. Read the report in order: the summary, then the sections on specific findings, then the research gaps. The gaps section is often the most useful part for anyone planning further research or writing a critical piece.
4. Click through to every paper you intend to cite. Deep reports are a map, not a substitute for reading.

4. Save and export

• Add useful papers to a Collection named for the project. Collections keep a literature search organised across sessions.
• When you are ready to write, export citations as RIS or BibTeX and import them into Zotero, EndNote or Mendeley, or export CSV for a spreadsheet-based screening log.

5. Connect it to your other tools (paid plans)

Pro and Deep include a monthly allowance of API and MCP uses. If you work in an AI assistant that supports MCP, connecting Consensus lets you pull cited research into a conversation without switching tabs. For scripted work, the API allowance covers the same searches programmatically. Usage beyond the included allowance is billed per use up to a monthly cap.

Tips to get better results

• Ask one question at a time. “Does X cause Y?” gives a clean Consensus Meter and a focused analysis. Compound questions produce muddled results.
• Specify the population. Evidence in athletes, older adults and children often points in different directions. Put the population in the question.
• Filter before you analyse. A Pro search over everything since 1980 is less useful than one over the last ten years of trials and reviews.
• Prefer reviews and trials for strong claims. Use the Study Snapshot to check study design before you cite a paper for a causal claim.
• Treat the Consensus Meter as a signal, not a verdict. It summarises a small set of top results. Open the papers behind each position.
• Spend Deep reviews on formed questions. Draft the question, run a Pro search first to check the literature exists, then commit a Deep review.
• Keep a Collection per project. It avoids re-running searches and keeps the papers you have already screened in one place.
• Check the full text when the abstract is thin. Snapshots and Ask Paper work from the text Consensus has. If a paper is paywalled, the available text may be limited, so follow the link to the publisher.

Limitations and common problems

• Coverage gaps. Consensus searches open scholarly indexes and its own crawl. Some paywalled or very recent papers may not be present or may have abstract-only text. Cross-check important searches in a library database.
• AI summaries can be wrong. Snapshots and analyses are generated. They occasionally misstate a sample size, direction of effect or population. Anything you cite must be checked against the paper.
• Monthly caps. The Free plan’s 10 Pro messages and 3 Deep reviews run out fast once you start using the tool seriously. Pro removes the Pro message cap but still limits Deep reviews to 15.
• Not a systematic review tool. There is no reproducible search strategy, screening workflow or risk-of-bias appraisal. Use it to scope and orient, then move to PubMed, Scopus or Web of Science for formal reviews.
• Yes/no framing. The Consensus Meter only appears for questions that can be framed as yes/no. Nuanced questions get an analysis without a meter.
• Monthly billing is more expensive. The displayed prices are annual-billing rates. Toggle to monthly on the pricing page to see what you will actually pay if you do not want a yearly commitment.

Alternatives

• Perplexity - a general answer engine with citations from the open web; better for current and non-academic questions, weaker on peer-reviewed depth. See our Perplexity vs ChatGPT for research comparison.
• Elicit - another AI research assistant aimed at literature review, with strong table extraction across papers.
• Semantic Scholar - a free academic search engine with AI summaries, no synthesis across papers.
• Google Scholar - the broadest free index, no AI analysis, best for exhaustive searching.
• Claude or ChatGPT - useful for drafting and critiquing once you have the sources; see our Claude review.

For more research tools, browse our AI research tools directory, and read our guide on how to write better prompts for question-writing techniques that carry over to Consensus.

Verdict

Consensus is worth signing up for if you regularly need to know what the research actually says and want sources you can defend. Start on the Free plan with three or four real questions; if you find yourself rationing Pro messages within a week, the Pro plan is cheap for the time it saves. The Deep plan is for people running literature reviews most weeks. Skip it if your questions are about the open web, current events or anything the peer-reviewed literature does not cover.