ASReview
Open-source active learning that reorders your screening pile as you work.
At a Glance
- Publisher
- Utrecht University
- Coverage
Not a database. A free, open-source active-learning framework for title and abstract screening in systematic reviews and evidence synthesis — you import records from your database searches and the model continuously re-ranks the unscreened remainder as you make decisions. Published benchmarks report recall around 96% with substantially reduced screening burden (see the peer-reviewed evaluations before relying on any specific figure).
- Access
Free and open source, installed locally via Python or run as a hosted instance by your institution. No vendor lock-in and your data stays where you put it.
- Free alternative
- Already free — Rayyan's free tier is the closest hosted alternative.
- Regional
Global; runs entirely on your own machine, which suits projects with data-residency constraints.
Our Take
ASReview is the rare evidence-synthesis tool that is both free and academically serious — it originated in a university methods group, its performance is documented in peer-reviewed benchmarks, and the code is open to inspection, which matters when a screening decision has to survive audit. The trade-off is setup: it expects some comfort with Python and it does not hold your hand through the wider review workflow the way Covidence or DistillerSR do. If you have the technical patience, it removes the licence cost from the most labour-intensive step of a review, and its stopping-rule question is a genuine methodological debate you should read up on rather than accept a default for.
Power Tips
- Seed the model with a handful of known-relevant and known-irrelevant records before you start — cold starts are slow.
- Decide and pre-register your stopping rule before screening; ‘until the model stops surfacing hits’ is not a defensible method.
- Export the full decision log — it’s your audit trail for the PRISMA flow diagram.
Alternatives & Complements
Rayyan, Covidence, DistillerSR, Nested Knowledge, Silvi.ai
Figures last verified: July 28, 2026