Revolutionize Your Literature Reviews

Transform weeks of manual screening into hours with AI. MetaScreener combines cutting-edge AI models with human expertise for systematic reviews.

95%*
Accuracy Rate
70%**
Time Reduction
10×***
Faster Processing
Get Started Now
Always free • AI-powered • Privacy-first

Why Choose MetaScreener?

Experience the perfect blend of AI efficiency and human expertise

Save Weeks of Work

Reduce systematic review screening time from weeks to hours. Our AI processes abstracts in seconds while maintaining research-grade accuracy.

Average time savings: 40-70%****

Research-Grade Accuracy

Achieve 95%+ sensitivity in identifying relevant studies with consistent application of your inclusion criteria across thousands of papers.

Cohen's Kappa: 0.85+ agreement*****

Human-AI Collaboration

Not a replacement for researchers - an enhancement. AI does the heavy lifting while you make the final decisions with full transparency.

Single reviewer with AI support******

Powerful Features

Everything you need for comprehensive literature screening and systematic reviews

Multiple LLM Support

Choose from OpenAI GPT, Anthropic Claude, Google Gemini, or DeepSeek models. Use your own API key for complete control over costs and data.

Multi-Format Processing

Process RIS files with abstracts or full-text PDFs with intelligent OCR. Extract text from scanned documents automatically.

Performance Analytics

Track screening performance with detailed metrics including confusion matrices, sensitivity, specificity, and Cohen's Kappa.

Data Privacy First

Your data stays secure. Processing happens locally in your browser with direct API calls using your own credentials.

Batch Processing

Process hundreds of documents simultaneously with multi-threading and real-time progress tracking for maximum efficiency.

Flexible Export

Export results in multiple formats: RIS for reference managers, CSV/Excel for analysis, JSON for custom processing.

Frequently Asked Questions

Everything you need to know about MetaScreener

Getting Started

No programming knowledge required. MetaScreener features an intuitive web interface that guides you through the entire process from configuration to results export.
MetaScreener supports RIS files (from PubMed, Scopus, Web of Science) and PDF documents for full-text screening. Our OCR technology handles scanned PDFs automatically.
Simply visit the configuration page, select your preferred AI provider, and enter your API key. We provide step-by-step guides for obtaining API keys from OpenAI, Anthropic, Google, and DeepSeek. Your keys are stored locally and never transmitted to our servers.

Costs & Pricing

MetaScreener is completely free. You only pay API providers directly for their services. We recommend DeepSeek for most cost-effective processing (typically $0.5-25 USD per review). No fees, commissions, or revenue sharing from our side.
DeepSeek offers the lowest costs while maintaining excellent quality. Cost comparison per 1000 abstracts: DeepSeek (~$1-3), OpenAI GPT-4o (~$8-15), Claude/Gemini (~$12-25). We provide cost calculators for accurate budgeting.

Performance & Accuracy

Based on validation studies with 4,230 articles from 18 institutions, MetaScreener demonstrates 95-97% sensitivity in identifying relevant studies. This means it rarely misses important papers. Specificity ranges from 85-92%, significantly reducing manual review workload.
MetaScreener enhances, not replaces, traditional methods. It accelerates the screening process while maintaining researcher oversight and decision-making authority. Final inclusion/exclusion decisions should always be validated by human experts.
MetaScreener provides detailed justifications for each decision, allowing you to review and override AI recommendations. The system is designed with high sensitivity to minimize false negatives (missing relevant papers), though this may increase false positives requiring manual review.

Technical & Features

Processing speed depends on the number of documents and chosen AI model. Typically: 100 abstracts in 2-5 minutes, 1000 abstracts in 20-40 minutes. PDF processing takes longer due to text extraction requirements.
Yes! MetaScreener allows you to define detailed, study-specific inclusion and exclusion criteria. The more specific your criteria, the better the AI can understand and apply your requirements consistently.
Your data is processed locally in your browser. API calls go directly from your browser to the AI provider using your credentials. We never store your documents, API keys, or processing results.

Support & Collaboration

MetaScreener works across all research disciplines. Our validation includes medical, social sciences, engineering, and public health domains. The AI models adapt to domain-specific terminology and research methodologies.
Yes, MetaScreener exports results in multiple formats including RIS files for reference managers (Zotero, Mendeley), CSV/Excel for data analysis, and detailed reports with AI justifications for team review.
We welcome testing contributions from research institutions worldwide! If you have systematic review datasets with gold-standard manual screening results, please contact us. Contributors will be acknowledged in our validation reports.
MetaScreener includes comprehensive documentation, video tutorials, and example workflows. For technical support, reach out through our GitHub repository or contact forms for assistance with setup and troubleshooting.

Ready to Transform Your Research?

Join researchers worldwide who are already using MetaScreener to accelerate their systematic reviews and meta-analyses.

Get Started Now
* Accuracy metrics based on internal testing with 2,000 research papers from public health domains, plus external validation data contributed by researchers from various institutions including University of Science and Technology of Macao, Peking University, University of Sydney, University of Wisconsin, University of Melbourne, Imperial College London, and others. Total contributed dataset: 4,230 articles spanning public health and clinical research areas. Sensitivity: 95-97%, Specificity: 85-92%. Performance may vary by research domain and inclusion criteria complexity.
** Time reduction data derived from internal efficiency testing and user feedback. Comparison between traditional dual-reviewer screening versus AI-assisted single-reviewer screening across test projects. Time savings typically range: 40-89%, average observed: 68%. Includes time tracking for literature screening, data extraction, and initial quality assessment phases.
*** Processing speed benchmarks conducted in controlled testing environments. Manual screening baseline: 150-300 abstracts/hour/reviewer. MetaScreener processing: 1,500-4,500 abstracts/hour, depending on selected AI model, document complexity, and network conditions. Benchmarks performed under standardized conditions with consistent hardware and software configurations.
**** Time savings analysis based on user surveys and usage data from research teams using MetaScreener. Calculation: (Traditional_Time - AI_Assisted_Time) / Traditional_Time × 100%. Includes screening, data extraction, and quality assessment phases. Results vary by domain: Medicine ~71%, Social Sciences ~64%, Engineering ~73%. Individual results may differ based on study complexity and user experience.
***** Inter-rater agreement measured using Cohen's kappa coefficient, based on available testing and validation data. AI-human agreement: κ=0.85-0.89, human-human agreement: κ=0.87-0.91. Test samples included varying complexity literature screening tasks. Results suggest AI-assisted screening maintains comparable reliability to traditional methods, though individual performance may vary.
****** Single-reviewer workflow evaluated through comparative testing across available systematic review projects. Primary outcome: inclusion/exclusion accuracy. Secondary outcomes: time efficiency, user satisfaction. Preliminary results suggest maintained accuracy (within 2% margin) with improved efficiency (~68% time reduction) and positive user feedback (average satisfaction: 8.1/10). Results based on limited sample size and may not generalize to all use cases.
Testing & Validation: Performance data sourced from internal testing environments and voluntary user contributions. Statistical analysis conducted using standard methods with α=0.05 significance level. Effect sizes calculated for continuous outcomes, odds ratios for categorical outcomes. Missing data handled appropriately. Results should be interpreted with caution given sample limitations. We encourage additional testing contributions from research institutions worldwide to improve validation robustness.
Quality Assurance: MetaScreener development follows standard software engineering practices including version control, automated testing, and code review processes. Algorithm performance monitored through user feedback and usage analytics. Regular updates and optimizations based on user suggestions and identified issues. No formal quality management certification or regulatory compliance claimed.
Pricing Information: MetaScreener is provided free of charge with no subscription fees, licensing costs, or hidden charges. Users pay AI service providers directly according to their published pricing models. Based on DeepSeek pricing (typically most cost-effective): Small reviews (≤500 papers): $0.5-2 USD, Medium reviews (500-2000 papers): $2-8 USD, Large reviews (2000+ papers): $8-25 USD. Costs vary significantly by provider and model version: DeepSeek (generally lowest), OpenAI GPT-4o (moderate), Claude/Gemini (typically higher). Actual costs depend on model selection, document length, processing complexity, and current provider pricing.