ASReview vs Colandr

ASReview is a free, open-source tool that uses AI-aided active learning to prioritize records for systematic review screening, with a human making screening decisions. Its site says the approach can reduce screening workload by up to 95%. Users can simulate AI-aided review to compare models, while Crowdscreen supports parallel screening with access managed by administrators through a private login system. The described model families are ELAS-Ultra, ELAS-Heavy, and language-agnostic ELAS-L; ELAS-Heavy and ELAS-L require the Dory package. ASReview works with reference managers including EndNote and PubMed, and imports and exports CSV, RIS, and XLSX files. Documentation includes a Python API, while the institutional workflow also describes a REST API and CLI for automation. In a local setup, data stays on the user's computer; server deployments keep it in the organization's environment and follow its policies. Support includes documentation, tutorials, videos, community discussion, and organizational workshops and consultancy. ASReview LAB costs nothing with no licensing fees or subscriptions, though running hardware or a server may incur costs.

Colandr is a free web platform for evidence reviews and other document reviews and syntheses. Machine learning ranks citations by relevance to help screen titles and abstracts, while natural language processing classifies included documents into user-defined categories and assists full-text data extraction. Reviewers retain the final inclusion, exclusion, and classification decisions, and teams can collaborate and cross-check work. Citation files can be imported in TXT, RIS, or BIB formats, with a 40 MB limit per import, and screening decisions and extracted data can be exported as CSV. Colandr does not connect directly to citation databases; users export citations elsewhere, although the platform can help create Boolean search strings. Full-text papers must be found and imported individually, with no mass-upload feature. Relevance rankings update after every 10 screened citations, and metadata suggestions begin after 50 hand-coded extractions. Colandr describes itself as open-source and open-access.