Screening that survives peer review
Journals expect two independent reviewers, a documented conflict process, and an agreement statistic. That is the default here: reviewers vote blind, neither sees the other's decisions, conflicts route to a resolver, and agreement reports as Cohen's kappa without anyone computing it by hand.
AI as the first pass, never the decision
Give the AI your inclusion and exclusion criteria and it scores every title and abstract for relevance, 0 to 100, with a recommendation and a one-line reason. The pile sorts by score so the obvious includes surface first and the obvious excludes sink. Nothing is decided for you: every include or exclude is a human click, which is what your methods section has to say anyway.
The unglamorous parts, handled
- RIS import from any database or reference manager, plus live PubMed and Europe PMC search inside the app
- Deduplication on DOI and PMID, with every duplicate pair shown for review before it is dropped
- A full-text stage tracking retrieval status, eligibility, and exclusion reasons
- PRISMA 2020 counts that fill themselves from your decisions
- RIS export of your final set for RevMan, EndNote, or Zotero
Then the review keeps moving
Included studies flow into data extraction, risk of bias, meta-analysis, GRADE, and a manuscript draft in the same free project. Screening tools that stop at screening leave you rebuilding the project somewhere else; this one does not stop.