Literature Intelligence
Search, analyze, and synthesize academic literature across multiple sources with AI-powered evidence extraction, citation network analysis, and document generation.
Overview
Literature Intelligence is a dedicated module within each CATO project that enables systematic literature search and analysis. It connects to seven academic data sources, uses five specialized AI agents to process results, and provides tools for evidence extraction, citation network visualization, and automated document generation.
Access Literature Intelligence from the Literature tab in any project. Results can be organized into collections, annotated, and sent directly to CATO Chat for further analysis.
Data Sources
Literature Intelligence searches across seven academic and patent databases simultaneously:
| Source | Coverage | Max Results |
|---|---|---|
| PubMed | Biomedical and life sciences journal articles | 10,000 |
| BioRxiv | Biology preprints with TF-IDF relevance ranking | 1,000 |
| arXiv | Physics, mathematics, computer science, and quantitative biology | 1,000 |
| ClinicalTrials.gov | Clinical trial registrations and results (v2 API with status/phase filters, automatic retry on failures) | 50 per search |
| Patents | Patent filings via USPTO Open Data Portal (automatic retry on transient failures) | 50 per search |
| OpenAlex | Citation metadata and cross-referencing | Varies |
| Semantic Scholar | Academic paper aggregation and citation analysis | Varies |
Query Studio
The Query Studio is where you compose and execute literature searches. Enter your research question in natural language, and the AI search planner agent will translate it into optimized queries for each data source.
Search Modes
Choose a search mode to optimize results for your specific research goal:
Evidence-First
Prioritizes clinical evidence, RCTs, meta-analyses, and systematic reviews. Best for clinical questions and evidence-based medicine.
Novelty-First
Surfaces the most recent publications and preprints. Best for staying current on rapidly evolving research areas.
Tip: You can select which data sources to include for each search. Deselect irrelevant sources to get more focused results and faster search times.
Study Type Filters
Filter search results by study design to focus on specific evidence types. Available study types include:
How Filtering Works
CATO uses two complementary filtering strategies depending on the data source:
- Server-side (PubMed) — Study types like RCT, Meta-analysis, Systematic Review, Review, and Case Report are filtered using native PubMed Publication Type [PT] tags. This is the most accurate method as filtering happens before results are returned.
- Post-filter (all sources) — All results are filtered using word-boundary matching on abstracts, venue, and type metadata. This uses regex patterns that prevent false matches (e.g., "RCT" will not match "extraction", "review" will not match "reviewed"). Hyphens, dashes, and spaces are treated as interchangeable for matching flexibility.
Note: Cohort and Case-control are not native PubMed Publication Types, so they rely on post-filter text matching only. For these types, results may be less complete than for types with server-side PubMed support.
Results Grid
Search results appear in a sortable, filterable grid with the following information for each paper:
- Title and authors — Full paper title with author list
- Source badge — Which database the result came from
- Relevance score — AI-computed relevance to your query
- Publication date — When the paper was published
- Abstract preview — Expandable abstract text
- DOI link — Direct link to the full paper when available
Select individual papers or use bulk selection to add them to collections, run evidence extraction, build citation networks, or send to Chat.
Evidence Matrix
The Evidence Matrix provides a structured, tabular view of key information extracted from selected papers by the AI evidence extractor agent. Each row represents one paper, and columns show:
| Column | Description |
|---|---|
| Relevance | AI-computed relevance score (0–100%) combining semantic similarity, evidence strength, recency, and source quality. Sortable; auto-sorted highest first by default. |
| Population | Study population characteristics (PICO: Population) |
| Intervention | Treatment or exposure being studied (PICO: Intervention) |
| Comparator | Control group or comparison (PICO: Comparator) |
| Outcome | Primary outcomes measured (PICO: Outcome) |
| Effect Size | Reported effect sizes (OR, HR, RR, etc.) with confidence intervals |
| Bias Indicators | Potential sources of bias identified by the AI |
The Evidence Matrix is particularly useful for systematic review workflows, allowing you to compare findings across studies at a glance.
Citation Network
The Citation Network visualizes relationships between papers as an interactive force-directed graph. Nodes represent papers and edges represent citation relationships.
Network Metrics
The graph builder agent computes several metrics to help you identify key papers:
- PageRank — Identifies the most influential papers in the network based on citation patterns
- Betweenness centrality — Finds papers that bridge different research communities
- Community detection — Groups papers into research clusters using the Louvain algorithm
- Hub & authority scores — Identifies papers that are well-cited (authorities) and papers that cite many key works (hubs)
Advanced Analysis
- Contradiction detection — Uses TF-IDF cosine similarity to identify papers with potentially conflicting conclusions
- Research gap identification — Analyzes network topology to suggest under-explored research areas
Tip: For meaningful network analysis, select at least 10-20 papers with DOIs. Papers without DOIs or reference metadata may not form citation edges.
Document Generation
The AI brief writer agent can generate structured documents from your selected papers and analysis results:
Systematic Review Brief
A structured summary following systematic review methodology with evidence synthesis, quality assessment, and conclusions.
Landscape Analysis
An overview of the research landscape including key themes, trends, major research groups, and emerging directions.
Gap Analysis Report
Identifies under-explored areas, methodological gaps, and opportunities for new research based on network topology and evidence analysis.
Generated documents are reviewed by the QA sentinel agent for factual accuracy before being delivered to you.
Library Management
The Library provides persistent storage for papers you want to keep across search sessions:
- Collections — Organize papers into named collections (e.g., "Background Reading", "Methods Papers")
- Annotations — Add personal notes and highlights to individual papers
- Reader Panel — View full abstracts, metadata, MeSH terms, and references in a dedicated panel
- Work Inspector — Detailed metadata view including DOI, PMID, authors, affiliations, and citation counts
- Tree & Grid views — Switch between hierarchical collection tree and flat grid layouts
Watch Alerts
Save any search query as a Watch Alert to automatically monitor for new publications. When new papers matching your criteria are published, you will be notified within the Literature Intelligence module.
- Run a search in Query Studio
- Click Save as Watch to create an alert
- New results will appear in your Watch results panel
- Review new papers and add relevant ones to your Library
Literature-to-Chat Bridge
Send curated sets of papers from Literature Intelligence directly to CATO Chat for AI-powered analysis. This bridges structured literature search with conversational AI capabilities.
How to Use
- Select papers from your search results or Library
- Click Send to Chat
- Choose a prompt type from the bridge dialog (see below)
- The Chat tab opens with the literature packet loaded and prompt template pre-filled
- Send the pre-filled prompt immediately or customize it before sending
Prompt Types
When sending papers to Chat, choose a prompt type that matches your intent:
Ask with context
Type your question and CATO answers using only the attached sources. The prompt template ends with a colon so you can immediately type your question.
Analyze and extract
CATO systematically analyzes the papers, extracting key findings, methods, and evidence. Ready to send immediately.
Draft a section
CATO writes a literature review section synthesizing the attached sources with inline citations. Ready to send immediately.
Plan next steps
CATO proposes follow-up research directions, identifies gaps, and suggests new experiments based on findings. Ready to send immediately.
Quick-Action Buttons
When papers arrive in Chat, quick-action buttons also appear for common tasks:
Summarize
Generate a concise synthesis of all sent papers
Find Conflicts
Identify contradictions and disagreements across papers
Propose Next Steps
Suggest follow-up research directions based on findings
Export Brief
Generate a downloadable DOCX summary document
Run Console
The Run Console provides real-time visibility into Literature Intelligence operations. When a search or analysis is running, you can monitor:
- Which AI agent is currently active and what it is processing
- Progress across data sources (e.g., "PubMed: 150 results, BioRxiv: 42 results")
- Token usage and processing time for each agent
- Any errors or warnings from individual connectors
AI Agents
Literature Intelligence uses five specialized AI agents, each handling a different stage of the research pipeline:
| Agent | Tier | Role |
|---|---|---|
| Search Planner | Fast | Translates natural-language queries into optimized search strategies per source |
| Evidence Extractor | Balanced | Extracts PICO elements, outcomes, effect sizes, and bias indicators from abstracts |
| Graph Builder | Fast | Constructs citation networks and computes graph metrics (PageRank, communities) |
| Brief Writer | Advanced | Generates structured documents (review briefs, landscape analyses, gap reports) |
| QA Sentinel | Fast | Validates generated documents for factual accuracy and citation correctness |
Systematic Review
The Systematic Review feature provides AI-powered multi-pass synthesis of large evidence sets across multiple studies. It includes:
- Narrative synthesis — Qualitative synthesis of findings across studies
- Study design detection — Automatic identification of study types from metadata
- Quality assessment — Evaluation of methodological rigor
- Evidence limitations — Notes on methodological weaknesses and bias risks
Tier Restriction: Systematic Review is available to Pro, Pay-as-you-go, and Enterprise users. Free and Base tier users will see an upgrade prompt when attempting to use this feature.
Example Workflows
Systematic Review Preparation
- Enter your PICO question in Query Studio (e.g., "immunotherapy vs chemotherapy for NSCLC overall survival")
- Select Evidence-First search mode
- Review results and select relevant RCTs and meta-analyses
- Run Evidence Matrix extraction to get structured PICO data
- Generate a Systematic Review Brief document
- Send the brief to Chat for further analysis or export as DOCX
Research Landscape Mapping
- Search for your research area using Novelty-First mode
- Select 20-50 key papers from the results
- Build a Citation Network to visualize research clusters
- Review community detection results to identify distinct research themes
- Check contradiction detection for conflicting findings
- Generate a Landscape Analysis document
Patent Landscape Search
- Use Prior-Art mode and enable the Patents source
- Search for your technology area
- Review patent filings alongside academic publications
- Build a network to see citation relationships between patents and papers
- Generate a Gap Analysis to identify unpatented research areas
Tips for Effective Searches
- • Use specific medical terms and gene/protein names for precise results
- • Choose the search mode that matches your research goal
- • Deselect irrelevant data sources to speed up searches
- • Save important papers to collections for long-term organization
- • Use Watch Alerts to stay current on active research topics
- • Select 10+ papers with DOIs for meaningful citation network analysis
- • Send curated paper sets to Chat for deeper AI-powered discussion
- • Use the Evidence Matrix to compare findings systematically before drawing conclusions