Title & Abstract Screening
Screen every collected paper against your abstract criteria, verify the AI decisions, and confirm the stage to move your review forward.
What is Title & Abstract Screening?
The first screening pass. Paperguide AI reads the title and abstract of every paper in your confirmed set, evaluates it against your criteria, and shows you the evidence behind every decision. A person then verifies that work before the review moves on.
How this stage works
- A Full Access member starts AI screening. AI evaluates every paper against your research question and abstract criteria.
- Results appear as they are produced. You can begin reviewing before the full run finishes.
- A verifier reviews the decisions. They check each one against the abstract and the evidence AI cited, and override anything that is wrong.
- The verifier confirms the papers. Only then does the next stage unlock.
Nothing advances on AI output alone. Verification is what moves the review forward.
Before you start
Set up your criteria
AI screens against the abstract screening criteria in your protocol. If they are not defined, this stage cannot run.
Criterion order matters. When a paper fails, the first criterion it fails, in the order you listed them, becomes its exclusion reason. Order your criteria so the exclusion reasons your review reports are the ones that actually matter.
Assign a verifier
Decide who is responsible for verifying this stage and assign them before screening starts. Assignment determines who can act on the results and who gets notified when the run completes.
| If the stage is | Who verifies it |
|---|---|
| Assigned to a member | That member. They receive the notification when screening finishes. |
| Not assigned to anyone | Any Full Access or Review member. All of them are notified and can work through the results together. |
View members can never verify, assigned or not.
Running AI screening
Only** Full Access**members can** start AI screening.** Review members verify results but cannot trigger the run.
Screening a large set takes time, but you do not have to wait for it. Results appear as papers are processed, and a progress banner tracks the run, so the verifier can start working through decisions while the rest are still being evaluated. Paperguide also sends an email when the run completes, to the assigned verifier or, if the stage is unassigned, to every Full Access and Review member.
How AI reaches a decision
AI evaluates each paper one criterion at a time, then derives a single overall decision from those results.
For every criterion, it returns a verdict and the evidence behind it:
| Verdict | What it means |
|---|---|
| Meets | The abstract satisfies this criterion |
| Fails | The abstract does not satisfy this criterion |
| Unclear | The abstract does not give enough information to judge |
Each verdict comes with cited evidence: the specific text from the abstract that AI based it on. This is what makes a decision checkable rather than something you either trust or do not.
The overall decision follows from those verdicts:
| If | Overall decision |
|---|---|
| Any criterion fails | Exclude, with that criterion as the exclusion reason |
| No criterion fails, but at least one is unclear | Maybe |
| Every criterion meets | Include |
Read the criterion verdicts before changing an overall decision. When you disagree with a result, the cause is almost always one criterion that was judged differently than you would have judged it, and that is the level to fix it at.
This logic keeps borderline papers in the review rather than dropping them early. A paper you cannot judge from its abstract becomes a Maybe and moves forward to be settled on its full text. In a systematic review, losing a relevant study is a more serious error than carrying an irrelevant one into the next stage.
Working through the results
Screening opens in a table view: papers on the left, abstract details on the right. Select a paper to open its detailed view, with the full abstract, the current decision, every criterion verdict, and the cited evidence.
Papers are grouped into four tabs so you can review by outcome: All, Included, Excluded, and Maybe. The Maybe tab is the one worth your attention first, since it holds the papers where AI could not reach a clean judgment and human reading matters most.
To manage a large queue:
- Filter by decision to work through one outcome at a time
- Filter by verification status to find papers that still need confirming
- Sort by title, year, citations, or decision date
Sorting by decision date and filtering to unconfirmed papers gives you a clean queue during an active run, with newly screened papers arriving at the end.
Verifying and overriding
The AI decision is a recommendation. A confirmed decision is one a person has reviewed and accepted.
To verify a paper, open it and check the criterion verdicts against the abstract, paying most attention to anything marked fails or unclear, since those drive the outcome. If you agree, confirm it. If you disagree, override the decision to Include, Maybe, or Exclude. When you override to Exclude, select which criterion the paper failed.
If the abstract genuinely does not support a clean decision, leave the paper as Maybe. That is the correct answer, not an avoided one.
To confirm, either confirm papers individually as you review them, or select Confirm All once you have worked through the set. Confirming records the AI decision as approved, or your override as the final decision. Both are kept, so the review holds a record of what AI decided and what a person changed.
Full Access members can review and override decisions after verification is complete, including decisions made by another member. This gives a review lead a way to correct the record without reopening the stage for everyone.
Confirming the stage
Once every paper is confirmed, you see the final counts: Included, Maybe, and Excluded. These feed the screening block of your PRISMA diagram.
Included and Maybe papers both move forward. Excluded papers stop here.
Confirming unlocks the next stage. Depending on how you set up the review, that is either Full Text Screening or, if you disabled full text screening, Data Extraction.
The next stage stays locked until this one is confirmed. If screening has finished but nothing is moving, check whether the papers still need confirming.
If your protocol or papers change
Screening can run again when the review changes underneath it.
If you change your abstract criteria in the protocol, screening may need to run again so decisions stay consistent with the criteria you are now using. How much needs rescreening depends on what changed. This is why the protocol is worth settling before you start.
If you add papers to the collection, only the new papers need screening. Work already confirmed is not affected, and a verifier reviews the new results the same way.
Adding papers after screening has started
Exporting screening data
You can export your abstract screening results, including decisions, exclusion reasons, and the evidence behind them. Useful for sharing progress with collaborators outside Paperguide and for keeping a screening record alongside your submission.