The Manual Screening Workflow
One paper at a time, with AI support at your side, not in the driver's seat.
The screening page is where you spend most of your time in a review. It's built around one paper at a time, with decision support alongside it rather than buried behind extra clicks.
The paper card
The center of the page shows the paper's title, authors, journal, volume, pages, year, a link to its DOI, and its abstract with any AI-highlighted evidence quotes. If enabled, a sidebar next to it shows the AI Suggestion (decision and reasoning) and the RI Score block (score, tier badge, the four sub-scores, and a short justification).
Making a decision
Three buttons: Include, Exclude, or Uncertain. Choosing Exclude opens a reason selector drawn from a standard list (e.g. wrong population, wrong study design, duplicate); if none of the fixed options fit, pick Other and type a short custom reason. You can add a free-text note to any decision. Once you're done, Save & Next moves you on, or Skip if you want to come back to a paper later, skipped papers are tracked so you can resume where you left off.
The left panel
Alongside the paper card, a panel tracks your progress: counts of Included/Excluded/Uncertain papers so far, and your agreement rate with the AI's suggestions. Tier filter chips (All, Uncertain, AI Included, AI Excluded) let you jump straight to the papers that need your attention most. A Run AI Screening button processes any papers that haven't been screened yet.
Bulk-accepting AI suggestions
If you trust the AI's recommendations for a batch of papers at a given tier, "Accept N included" or "Accept N excluded" applies those recommendations as your final decisions in one action. Because this skips individual review, each one shows a confirmation dialog first, with a "don't show this to me again" option if you'd rather not see it every time. Bulk-accepted decisions are recorded in your audit trail exactly like individually reviewed ones, and you can still change any of them afterward from the Papers page.
The Inclusion Threshold
A slider lets you set where the RI score's include/exclude boundary sits for your review, see Relativity Index Scoring for how it works and what moving it actually changes.
Contradiction flags
If you make opposite decisions (one include, one exclude) on two papers that landed in the same RI tier, Zynthia flags it as a possible contradiction, an amber banner links you to the other paper so you can compare them. If the difference is intentional (the papers really do differ in a way the tier doesn't capture), you can override the flag and note that it's correct; otherwise, it's a nudge to reconsider one of the two decisions before it undermines your review's consistency.