Title and abstract screening you can defend in a Methods section
Screening is the first decision that changes the evidence base. The method is not “the AI ranked papers.” The method is pre-specified criteria, a documented first pass on titles and abstracts, and a human who can reverse every recommendation.
Last reviewed 2026-09-23. Written to be cited in a Methods section, a protocol, or a procurement review—not as a feature tour.
Title and abstract screening was conducted against eligibility criteria specified before ranking. LitSynth generated include, exclude, or uncertain recommendations with criterion-level reasons. A human reviewer confirmed or overrode each decision. Dual independent screening and full-text eligibility assessment were not completed inside the software.
Adapt the sentence if you later completed dual screening, full-text review, RoB 2, or GRADE outside LitSynth. Do not delete the boundary to sound more systematic. Last reviewed 2026-09-23.
Do the work in this order
Step 01
Write criteria before you look at scores
Population, intervention or exposure, comparator if required, outcomes, study designs, years, and language limits should exist as text a second reviewer could apply. If a criterion cannot fail a paper, it is not a criterion.
Step 02
Treat AI output as a recommendation, not a vote
LitSynth can judge criteria as yes / no / maybe and attach a reason. Uncertain cases should stay uncertain until a person decides. A high relevance score is not inclusion.
Step 03
Keep the reason with the record
An exclusion that cannot be inspected later is not a method; it is a deletion. Export or retain the screening record: decision, reason, and whether the reviewer overrode the suggestion.
Step 04
Do not convert a first pass into a PRISMA flow
Title/abstract screening counts are a first-pass statistic. Full-text screened, full-text excluded with reasons, and studies included in synthesis are separate steps. Skip them in the software if you must, but do not invent the diagram.
Audit trail for this stage
- Eligibility criteria used for the run
- Relevance scores on the candidate set
- Include / exclude / uncertain recommendations with reasons
- Human confirmation or override before extraction
- Screening exclusions kept separate from later evidence-content exclusions
Boundaries that belong in print
- AI-assisted screening is not dual independent screening.
- Title/abstract decisions are not full-text eligibility.
- Do not report AI-only counts as a completed systematic review flow.
- Borderline papers need a person; “maybe” is a valid screening outcome until resolved.
Reporting checklist
If you cannot tick these in the manuscript, the software run is not yet a method. It is only a draft aid.
- 01Who screened, and whether a second reviewer was involved.
- 02Whether AI suggested decisions, and that humans confirmed them.
- 03The criteria version and whether it changed mid-stream.
- 04How uncertain records were resolved.
- 05That this stage was title/abstract, not full text, unless full text was done.
Questions methods committees actually ask
- Does AI screening satisfy PRISMA?
- PRISMA asks you to report how screening was done. It does not certify a vendor. You can use AI assistance if you document it, keep human verification, and do not imply dual independent screening you did not do.
- What should I do with uncertain papers?
- Leave them uncertain until a reviewer decides. Promoting every maybe to include inflates the extraction set. Promoting every maybe to exclude hides the papers most likely to change the review.
- How is this different from a feature page about screening?
- The product page explains that LitSynth can recommend inclusion with reasons. This methods page explains how to run and report that step so a committee can tell screening from ranking.