Contains over 220 million papers across scientific domains
Primary sources: Semantic Scholar, OpenAlex, plus proprietary web crawling. Also incorporates data from PubMed, Crossref, ORCID, etc.
Updates weekly to maintain currency
Combines multiple search technologies:
Uses a three-stage hybrid search (semantic AI and BM25), custom fine-tuned models for specialized tasks, and proprietary relevance-checking models to reduce hallucinations
MeSH synonym support for medical queries
Ranks papers by relevance, citation count and velocity, study design, and publication date
User Interaction:
Search using questions, keywords, or DOI
Supports Boolean operators (AND, OR, NOT)
Apply filters for:
Publication date
Preprint exclusion
Open access
Minimum citations
Methods
Journals
Subject domains
Country
Study duration
Sample size
Population type (human, animal, in vitro)
Study types (meta-analyses, systematic reviews, RCTs)
Export results to CSV or RIS with detailed paper-level data. Save and organize papers with bookmarks and custom lists (unlimited in Premium, one in Free)
Analysis Features:
Consensus Meter (Version 2.0):
Analyzes yes/no questions across top 20 results
Classifies results as "yes," "no," "possibly," or "mixed"
Shows percentage of agreement across papers
Displays quality indicators (methodology, recency, journal prestige, citations), with badges highlighting which positions score highest in each
Study Snapshot:
Displays up to 7 categories of information
Includes population, sample size, location, results, outcomes, duration, and methods
AI-powered extraction from paper abstracts
Accessible via sidebar view for seamless workflow
Pro Analysis (formerly Copilot):
Synthesizes up to 20 papers using full-text (when available), with TL;DR headers, structured comparisons, and in-line citations in multiple formats.
Supports 31+ languages
Creates claims & evidence tables and research gaps heat maps
"Ask Paper" feature (beta) for full-text paper interactions:
Chat with individual papers to ask clarifying questions
Works with papers from the corpus and user-uploaded PDFs
AI research agent that conducts comprehensive literature reviews
Breaks questions into sub-questions and runs up to 20 targeted searches
Reviews over 1000 papers, selects the ~50 most relevant, and generates cited reports with visuals, consensus analysis, citation graphs, and key author insights
Completes in ~2 minutes with email notification
Free: 3/month \ Pro: 15/month | Deep: 200/month
Threads:
Conversational search that maintains context across queries
Colored citations showing stance (Yes/No/Maybe)
Thread Citations panel with sortable paper list
Smart follow-up prompt suggestions
Research Hub:
Upload and analyze your own PDF documents
Use Ask Paper features on personal research papers
Table View:
Side-by-side paper comparison
Extracts study attributes (population, methods, results, sample size, duration)
Access via "Table" button in search results
Notable Features:
Supports auto-citations via Zotero, Mendeley, EndNote, and RefWorks. Offers customizable UI (Dark/Compact modes) and 100+ language support. Multiple export formats available (CSV, RIS).
Quality indicators:
Rigorous Journal badge (top 50% on SciScore Index)
Highly Cited and Very Highly Cited badges
Study type tags (RCT, Systematic Review, Meta-Analysis, Literature Review)
Influential Citations indicator
Full-text availability indicator
ConsensusGPT integration in ChatGPT (requires ChatGPT Plus)
API access for embedding Consensus in custom tools
Research quality filters
Benefits:
Evidence-based scientific coverage
Direct connection to source papers
Multilingual content support
Detailed paper-level insights
Limitations:
Freemium model with limited free features:
Free: 25 Pro Analyses, 3 Deep Searches, 10 Study Snapshots, 10 Ask Paper messages per month
Pro: ($15/month): Unlimited analysis, 15 Deep Searches
Deep ($65/month): 200 Deep Searches plus all Pro features
Teams and Enterprise plans available for organizations
Search draws from multiple major databases, but niche fields may have gaps. Relevance is prioritized over recency, though publication date is a ranking factor. Some AI model details are proprietary. Advanced features like Deep Search require paid plans.
Can produce some off-topic suggestions despite relevance checking
