Consensus

Learning Objectives

• Understand what Consensus does and why specialized academic search exists alongside general AI tools
• Identify the key features: claim extraction, Consensus Meter, paper summaries, and citation export
• Evaluate when to use Consensus vs. Elicit, Google Scholar, or general AI research tools

What Is Consensus?

Consensus is an AI search engine purpose-built for scientific literature, founded in 2022 and backed by Y Combinator. Rather than searching the web broadly, Consensus searches a database of over 200 million peer-reviewed research papers and uses AI to:

1. Find papers relevant to your question
2. Extract the key empirical claim from each paper
3. Synthesize across papers to determine what the evidence says collectively — the "consensus"

The core value proposition: when you want to know whether something is supported by scientific research, Consensus gives you a sourced, evidence-based answer rather than an opinion. "Does intermittent fasting improve metabolic health?" returns a synthesis across dozens of relevant studies, not a general AI answer.

Core Features

The Consensus Meter

For questions with a clear empirical answer, Consensus displays a Consensus Meter — a visual indicator showing what percentage of the relevant papers found support, oppose, or are inconclusive about the claim.

• Example: "Does exercise reduce depression symptoms?" → Consensus Meter shows high support, moderate support, or mixed results based on extracted findings
• The meter is only shown when there is sufficient evidence and clear directional findings — ambiguous topics show a breakdown instead
• Each position in the meter links to the papers supporting that finding

This feature is genuinely useful for quickly understanding whether a claim is well-supported, contested, or uncertain in the literature — without reading every paper.

Claim Extraction and Paper Summaries

For each relevant paper, Consensus uses AI to extract:

• The core empirical claim the paper makes
• The study type (randomized controlled trial, meta-analysis, observational, etc.)
• Population studied, intervention used, and outcome measured
• Whether the finding was statistically significant

This structured extraction allows you to quickly assess paper quality and relevance without reading abstracts manually.

Synthesis Response

For premium users, Consensus generates a Synthesis — a written summary of what the evidence collectively says, structured like an answer with citations:

"Based on 23 papers reviewed, there is strong evidence that [X] correlates with [Y], with 18 of 23 papers reporting statistically significant effects. Most studies were conducted on [population], and the strongest evidence comes from [RCTs/meta-analyses]. Notable limitations include..."

Citation Export

Export search results and citations directly to:

• Zotero
• Mendeley
• CSV / Excel
• BibTeX for LaTeX documents

This makes Consensus a practical starting point for academic writing workflows.

Study Type and Quality Filters

Filter results by:

• Study type: RCT, meta-analysis, systematic review, cohort study, case study
• Year: Focus on recent research
• Journal/venue: Prioritize high-impact publications
• Sample size: Filter for adequately powered studies

Pricing

Strengths

• Evidence quality: Only searches peer-reviewed literature — not blogs, news, or opinion content
• Consensus Meter: Visual scientific consensus indicator is immediately useful for health, wellness, and empirical questions
• No hallucination risk for citations: Every citation is a real, retrievable paper
• Structured claim extraction: See what each paper actually says without reading the full abstract
• Citation export: Practical for researchers building paper databases
• Study type filtering: Prioritize meta-analyses and RCTs over weaker evidence

Limitations & Considerations

• Academic papers only: Cannot answer questions without published research — newer AI developments may not be covered
• English bias: Primarily English-language literature
• Free tier limits: 20 searches/day is adequate for light use but restrictive for heavy researchers
• Publication lag: The latest research takes months to years to appear in peer-reviewed journals — for fast-moving topics, literature may be behind
• Requires critical reading: Consensus Meter and AI summaries are starting points — important decisions should always involve reading the full papers
• Niche topics: Very specialized or emerging topics may have insufficient literature for meaningful consensus

Best Use Cases

When to choose alternatives:

• Comprehensive academic database searching → PubMed, Google Scholar, Scopus
• Broader web research → ChatGPT Deep Research or Perplexity
• Research against your own documents → NotebookLM
• More granular paper analysis → Elicit (stronger extraction; experimental trial filters)
• Real-time web + academic mix → Perplexity with academic sources enabled

Getting Started

1. Go to consensus.app — no account required for a few searches
2. Type a research question in natural language: "Does meditation reduce anxiety?" or "What are the effects of sleep deprivation on cognitive performance?"
3. Review the Consensus Meter (if shown) and read the extracted claims below
4. Click any paper to see its abstract and access the full text (via DOI or open access link)
5. For a research project, create a free account and use citation export to build your reference library

Key Takeaways

• Consensus searches 200+ million peer-reviewed papers and extracts empirical claims to show what scientific literature actually supports
• The Consensus Meter provides a visual overview of the evidence direction — high, moderate, or mixed support — for empirical questions
• Claim extraction shows what each paper actually concluded without requiring you to read every abstract
• Best for health, psychology, and social science questions that have a body of peer-reviewed research — not useful for very recent events or non-empirical questions
• Pair with Elicit for deeper paper analysis, and with Perplexity or ChatGPT for broader research that mixes academic and non-academic sources