A systematic search is only as credible as the trail it leaves behind. When a coauthor asks why a paper was excluded, when a reviewer requests the full search string, or when a new lab member inherits a review, the documentation is what lets the method survive the people who built it. This guide maps the decisions and artifacts that make a search auditable, from the first eligibility criteria to the final rerun before submission.
What does a complete search record actually contain?
Cochrane's guidance frames search planning around eligibility criteria: the search is built to find studies that could meet the review's inclusion rules, drawing on multiple sources and combining free-text terms with subject headings. The record you keep should make each of those choices visible. In practice that means a set of linked artifacts rather than a single document:
- Eligibility criteria written before searching, so the query has a stated target.
- The full query strings for every database and source, including field tags, filters, and date limits.
- A source list showing which databases, registers, and supplementary sources were searched.
- A dated log of when each search ran and what it returned.
- Screening decisions with reasons, tied to individual records.
- A study-linking map connecting multiple reports of the same study.
- A rerun record showing searches were refreshed before publication.
PRISMA 2020 supplies the reporting backbone for this material: a checklist, expanded explanations, and flow-diagram templates for original and updated reviews, all aimed at transparent reporting of search, selection, and synthesis methods. Treat PRISMA as the shape of the final report and the artifacts above as the raw material that fills it.
Where are you entering the workflow?
Readers arrive at search documentation from very different starting points. Diagnose your state before choosing a route.
| Your situation | What already exists | First move |
|---|---|---|
| Revising a manuscript after peer review | A search was run, but the record is thin | Reconstruct queries and screening reasons from database histories and notes |
| Closing a project | Searches and screening are done | Consolidate artifacts into a PRISMA-aligned package and rerun searches |
| Preserving an evidence trail for handoff | Partial logs scattered across tools | Standardize the log format and link records to decisions |
| Starting a new review | Nothing yet | Write eligibility criteria first, then build queries from them |
Each route converges on the same destination: a record another researcher could follow without asking you questions.
How do eligibility criteria shape the query?
Cochrane's approach starts searches from eligibility criteria rather than from a topic phrase. That ordering matters because it forces you to name the populations, interventions, comparators, outcomes, and study designs the review will accept before you type a single term. The criteria then generate the concept blocks of your query.
A practical sequence:
- Write the inclusion and exclusion rules in plain language.
- Break the rules into concepts, and list synonyms and variant spellings for each.
- Identify subject headings available in each database for those concepts.
- Combine free-text terms with subject headings so the query catches both indexed and unindexed records.
- Record the criteria and the resulting query side by side, so the link between them is visible.
When you search through WisPaper's scholarly search interface, keep the criteria document open beside it. The query you build there should be traceable back to a specific criterion, and any broadening or narrowing should be noted with a reason.
How do you document queries across multiple sources?
Cochrane recommends searching multiple sources, which means your documentation has to handle more than one query syntax. PubMed, for example, supports field tags, phrase searching, filters, Advanced Search, and saved searches, and its search details show how terms were translated and combined. That translation view is valuable documentation: it captures what the database actually did with your input, not just what you typed.
For each source, record:
- The exact string as entered, including tags and filters.
- The date the search ran.
- The number of records returned.
- Any interface settings that affect results, such as date limits or language filters.
- A note on how the string was adapted from the master query.
Google Scholar's documented features give you additional levers to log: searching by author and title, date sorting and limits, Cited by and Related articles, citation export, and email alerts. If you use alerts to keep a search current, note the alert query and its creation date so the ongoing surveillance is part of the record rather than an invisible habit.
How should screening be staged and recorded?
Cochrane describes screening in stages, and each stage produces its own decision record. A common structure moves from title and abstract review to full-text assessment, with reasons captured at every exclusion.
| Stage | Decision recorded | Artifact |
|---|---|---|
| Deduplication | Which duplicate was retained | Deduplicated record list |
| Title and abstract | Include, exclude, or uncertain | Screening sheet with reason codes |
| Full text | Include or exclude | Reason per excluded report |
| Study linking | Which reports belong to one study | Study-to-report map |
Deduplication deserves its own line because Cochrane calls for it explicitly, and because the count of duplicates removed feeds directly into the PRISMA flow diagram. Study linking matters for the same reason: when multiple reports describe one study, the flow diagram and synthesis both depend on counting studies rather than reports.
Use consistent reason codes so that a later reader can tally exclusions without re-reading every note. If two reviewers screen independently, record who screened what and how disagreements were resolved.
How do you keep the evidence trail organized as it grows?
A search record becomes hard to manage once records, PDFs, and notes accumulate. Organizing papers in a personal library lets you save or upload papers and group them by screening status, so the included set stays distinct from the pending and excluded sets. Pair that with a Research Projects area when the review is one workstream among several and you want the search record to sit alongside the rest of the project.
For questions that arise during screening, asking against the papers you have already collected in Library QA can help you check how a candidate paper treats a concept before you commit to an inclusion decision. Keep the answers as notes attached to the record, not as replacements for reading the paper.
How do you verify references before they enter the record?
Reference lists are a frequent source of error in a search trail. A citation that looks plausible can be wrong in its authors, year, or venue, and once it enters your included set the mistake propagates into the manuscript. TrueCite checks BibTeX against real academic databases to flag hallucinated references, which makes it a useful gate before references are added to the record. Run your BibTeX through TrueCite at the point where you are consolidating references, and log the check alongside the screening decisions.
How do you rerun searches before publication?
Cochrane recommends rerunning searches before publication, and PRISMA 2020 provides templates for updated reviews as well as original ones. The rerun is not a formality: it is a dated event that belongs in the record.
A rerun entry should state:
- The date the refresh ran.
- The exact strings used, which may differ from the original if criteria changed.
- The number of new records retrieved.
- The screening outcome for each new record.
- Whether the flow diagram counts were updated.
If your original searches used saved searches or alerts, the rerun can reuse them, but record that you did so and confirm the saved query still matches the current criteria. When criteria change mid-review, note the change and its date so the reader can see which records were screened under which rules.
What does a handoff-ready package look like?
A package is ready when a colleague can reconstruct your search without interviewing you. Test it against these questions:
- Can someone rerun each query exactly as you ran it?
- Does every excluded record have a reason?
- Are multiple reports of the same study linked?
- Do the flow-diagram counts reconcile with the screening sheets?
- Is the date of the final rerun recorded?
- Are references verified against a database?
If any answer is no, the gap tells you which artifact to rebuild. For longer consolidation work, a Scholar Agent task can support literature review work within WisPaper's agent area, but the decisions and their reasons remain yours to record.
How do you know the documentation is finished?
Finished documentation satisfies three tests. First, it is reproducible: the queries and settings are complete enough to rerun. Second, it is reconcilable: the counts in the flow diagram match the screening sheets and the deduplication log. Third, it is attributable: each decision has a person, a date, and a reason attached.
Run a final read-through as if you were the reviewer. If you cannot trace a number in the flow diagram back to a record, fix the trail before submission. PRISMA's expanded explanations are a useful checklist for this pass, since they describe what transparent reporting of search and selection methods should include.




