From research question to published evidence.
AutoEvidence handles the screening, extraction, and statistical synthesis that consumes months of researcher time — so you can focus on the science.
AutoEvidence by the numbers
One systematic review, end-to-end
Watch a real GLP-1 receptor agonist meta-analysis flow through all five steps. The same workflow that runs in production, replayed in 30 seconds.
Adults with type 2 diabetes (≥18 years)
GLP-1 RAs (semaglutide, liraglutide, dulaglutide)
Placebo or standard care
HbA1c reduction, MACE, all-cause mortality
Hover to pause · click any step to jump
Everything a systematic review demands.
A complete evidence-synthesis workflow from PICO to publication, in one platform.
Dual-pass AI screening
Claude AI screens every abstract twice and flags borderline cases. Cohen's κ inter-rater reliability scores tell you exactly where to focus.
Meta-analysis engine
Random- and fixed-effects models with forest plots, funnel plots, and I² heterogeneity — computed in-platform, exportable to your manuscript.
Multi-database search
PubMed, Embase, Cochrane, and ClinicalTrials.gov queried simultaneously with one PICO.
Vector deduplication
Fuzzy matching plus embedding similarity eliminate duplicates before screening.
Risk of bias
RoB2 and ROBINS-I frameworks with auto-generated GRADE evidence tables.
Protocol generation
PRISMA-P compliant protocols with eligibility criteria, amendment audit trail, and PROSPERO record import.
Full-text review
Structured screening with exclusion reasons mapped to PRISMA flow categories.
PRISMA 2020 reporting
Auto-generated PRISMA flow diagrams and full reports — exportable as JSON, CSV, RIS, BibTeX, or SVG.
From question to evidence in five steps.
A guided workflow that matches how systematic reviews are actually conducted.
- 01
Define your PICO
Set population, intervention, comparator, and outcome with guided templates.
- 02
Search & import
Query multiple databases. We deduplicate and import in seconds.
- 03
AI screens abstracts
Claude processes thousands of abstracts. You review only the borderlines.
- 04
Extract & assess
Structured data extraction and automated risk of bias assessment.
- 05
Meta-analyse & report
Statistical synthesis, visualisations, and PRISMA-compliant export.
Frequently asked questions
How the platform works, what it automates, and where you stay in control.
What is AutoEvidence?
AutoEvidence is an AI-accelerated platform for systematic reviews and meta-analyses. The AI handles the screening, extraction, and statistical synthesis that consumes months of researcher time, while you keep the scientific judgment — from PICO definition to a PRISMA 2020-compliant report.
How accurate is the AI screening?
Every abstract is screened twice (dual-pass) and borderline cases are flagged for human review, reaching 98% screening accuracy. Cohen's κ inter-rater reliability scores show you exactly where the AI reviewers disagree, so you focus your time where it matters.
Which databases does AutoEvidence search?
PubMed, Embase, Cochrane, and ClinicalTrials.gov are queried simultaneously from a single PICO. Fuzzy matching plus vector-embedding similarity removes duplicates before screening starts.
Is AutoEvidence PRISMA compliant?
Yes. Protocols are generated PRISMA-P compliant with an amendment audit trail and PROSPERO record import, full-text exclusion reasons map to PRISMA flow categories, and PRISMA 2020 flow diagrams and reports are generated automatically.
Can I run a meta-analysis directly in the platform?
Yes. Random- and fixed-effects models with forest plots, funnel plots, and I² heterogeneity are computed in-platform, alongside RoB2 and ROBINS-I risk-of-bias assessment and auto-generated GRADE evidence tables.
What export formats are supported?
Reports, extraction tables, and reference lists export as JSON, CSV, RIS, BibTeX, or SVG — ready for your manuscript or reference manager.