Consensus provides question-led search and AI-assisted synthesis of scientific research. Citation-backed output is useful only when the cited paper supports the claim and its methods fit the question. This review is documentation-based; we have not conducted an independent search-coverage or answer-accuracy benchmark.
Evidence updated 14 October 2026. Vendor statements are attributed to the linked primary sources. Suggested evaluation steps are not completed tests.
Search and synthesis are different tasks
The current search guide distinguishes paper search from modes that analyse or synthesise research. Use paper search to examine individual results, then verify central claims in the original studies. An automated consensus indicator is not a clinical recommendation, an assessment of every relevant paper or proof that all studies have comparable quality.
Pricing and allowances
Consensus’ product changelog records pricing and feature changes. Old launch posts describe beta-era limits that should not be repeated as current plan allowances. We have removed the previous fixed-limit table and universal “only peer-reviewed” guarantee; check the current plan and each paper’s publication status directly. Search access, synthesis modes and API usage may have different limits or billing units.
A repeatable evaluation
Choose a narrow question with several known studies and include a question on which the evidence is mixed. Record the query, date, returned papers and generated summary. Check whether population, intervention, outcome and uncertainty are preserved. Look for an omitted contradictory result and inspect whether a citation actually supports the sentence attached to it. This is a recommended evaluation, not a set of test results from this publication.
Verdict
Consider Consensus for discovering research around a question and producing an initial reading list. Use it alongside direct paper reading and a suitable literature-search method. ResearchRabbit offers a different starting point when citation networks matter more than question-led synthesis. We give no numeric score, “gold standard” label or guaranteed factual accuracy without a documented comparison. In high-stakes use, the tool’s summary alone is insufficient evidence for a decision.
Primary sources and related reading
Compare the decision with ResearchRabbit, Cursor and see how we review AI tools or our practical tutorials.