Consensus
An AI search engine that answers research questions using 200M+ real academic papers.
Overview
Consensus is built for researchers, students, and anyone who needs an evidence-based answer instead of a generic web search result. Ask a research question, and it searches 200M+ academic papers first, then uses AI to synthesize what the literature actually says, with citations back to the source.
Its "search first, synthesize second" design is meant to prevent the AI from inventing sources, a real risk with general-purpose chatbots. The tradeoff worth knowing about: the Consensus Meter's Yes/No/Possibly framing is a simplification that some in the research community consider too blunt for nuanced evidence synthesis.
Key Features
Hybrid Search: Combines semantic AI embeddings with keyword (BM25) search for relevance and precision.
Consensus Meter: A Yes/No/Possibly breakdown of how the literature answers your question.
Deep Search: Reads dozens of full papers and produces a structured, cited report.
Chat With Papers: Ask a specific paper follow-up questions directly.
Quality Ranking: Ranks results by recency, citation count, and journal impact factor.
LibKey Integration: Institutional access to paywalled articles.
Pricing
Starting price
$10/mo
Free: $0/month. Limited Pro Searches and Study Snapshots each month.
Premium: $15/month ($10/month billed yearly). Unlimited Pro Searches plus 15 Deep Searches a month.
Teams/Enterprise: Custom pricing for institutional and library-system access.
Note: consensus.app blocked direct automated verification (HTTP 403); figures corroborated across multiple independent sources. Confirm before buying.
Pros
Search-First Design: Searches real academic papers before synthesizing, reducing the risk of fabricated sources.
At-a-Glance Answers: The Consensus Meter gives a fast read on what the evidence says.
Deep Search Mode: Reads dozens of full papers into one structured, cited report.
Generous Free Tier: 10 Pro Analyses and 10 Study Snapshots a month, free.
Academic Reputation: Used across university library systems.
Cons
Oversimplified Framing: The Yes/No/Possibly meter can be too blunt for nuanced research questions, a real critique from the research community.
Still Fallible: AI-generated classifications can still be wrong, even with the search-first design.
Affiliate Program Unclear: consensus.app blocked direct verification, and search results conflate it with an unrelated same-named company (goconsensus.com).
Deep Search Is Paid: Deep Search and higher usage tiers require a paid plan.
What Makes It Unique
Literature-First, Not Model-First: Searches the literature before generating any AI summary, rather than asking an AI to answer from its own training data, the main defense against fabricated citations.
Kay Score
7
/ 10
Tool Information
Pricing
$10/mo
Category
Writing & Content
Platform
Web / iOS / Android
Last Updated
