Every Step. Automated.
From search strategy to publication-ready manuscript, ResBot handles the entire systematic review pipeline with AI precision and human oversight.
AI Search Strategy
Input your research question or PICO elements. AI generates optimized PubMed Boolean search strategies with MeSH term suggestions. Search PubMed, Scopus, OpenAlex, and more from a single interface.
- PICO-based query generation
- MeSH term auto-suggestion
- Multi-database search (PubMed, Scopus, OpenAlex)
- Search strategy validation
- Query expansion and refinement
Smart Screening
AI screens thousands of titles and abstracts in minutes using batched concurrent processing. Chain review catches errors. Manual override always available.
- 12 references per batch, 10 concurrent streams
- High-recall policy: defaults to INCLUDE
- Chain review for low-confidence decisions
- Keyboard shortcuts (j/k, i/e)
- Conflict resolution for multi-reviewer
Full-Text Pipeline
Automatic PDF retrieval from 15+ academic sources. AI OCR recovery for scanned documents. Dual-pass screening with consensus adjudication.
- Auto-retrieve from Unpaywall, PMC, CrossRef, CORE, and more
- AI OCR recovery for scanned PDFs
- Section detection (methods, results, conclusions)
- Dual-pass AI screening with consensus
- Built-in PDF viewer with annotations
Data Extraction Studio
Adaptive multi-pass extraction from text, tables, and figures. AI discovers the optimal extraction schema from your first studies, then extracts systematically.
- Schema discovery from first 3-5 studies
- 24 extraction types supported
- Evidence quality gates per field
- Cross-study normalization
- Side-by-side PDF + extraction form
Risk of Bias Assessment
10 validated RoB tools with AI-assisted domain assessment. Every judgment links to the evidence passage in the PDF. Traffic light matrix and robvis export.
- ROB 2, ROBINS-I, NOS, JBI, QUADAS-2, AMSTAR-2
- ROBINS-E, QUADAS-C, PROBAST+AI, Custom tools
- Signalling question algorithms
- Click judgment → highlighted PDF passage
- Traffic light matrix and weighted summary
Meta-Analysis Engine
Complete deterministic statistical engine. 11 effect measures, 7 tau² estimators, HAKN correction, publication bias tests, and 13 interactive chart types.
- OR, RR, RD, MD, SMD, HR, proportions, correlations, DOR
- DerSimonian-Laird, REML, PM, and more estimators
- Interactive forest plots with click-to-provenance
- Funnel plots with Egger regression
- R code generation (meta/metafor packages)
Manuscript Generation
AI generates publication-quality manuscripts in the correct order: Methods → Results → Discussion → Introduction → Abstract → Title. Real citations from PubMed and CrossRef.
- 6-section generation with persona-based prompts
- Real PubMed/CrossRef citation search
- PRISMA 2020 compliance (27/27 items)
- DOCX export (Times New Roman, double-spaced)
- LaTeX, XLSX, and supplementary materials
What Sets ResBot Apart
4-Tier AI Pipeline
Not one model fits all. Fast screening with Gemini Flash, precise extraction with GLM-5, independent chain review with MiniMax. Each task uses the optimal model.
Chain Review System
A second AI model independently verifies every decision. Catches fabrication, accuracy errors, and missed information before they reach your manuscript.
15+ PDF Sources
Unpaywall, PMC, CrossRef, Semantic Scholar, OpenAlex, Europe PMC, BioRxiv, CORE, DOAJ, Internet Archive, and more. If a PDF exists, we find it.
Auditable Everything
Full provenance trail. Every AI decision links to source evidence. Every screening choice is logged. Every extraction has a confidence score.
Real Math, Not AI Guessing
The statistical engine is deterministic: DerSimonian-Laird random-effects, HAKN correction, Knapp-Hartung. No AI hallucinated p-values.
Human-in-the-Loop
AI assists, humans decide. Manual override at every step. Blind screening mode. Multi-reviewer with inter-rater reliability calculation.