A retraction check is a verification task, not a single lookup. The reliable answer comes from combining several sources, recording what each one showed on the day you checked, and keeping that record with the manuscript or project it affects. This guide walks through a complete, auditable sequence you can run before citing a paper, during peer review, at a lab handoff, or when you reopen a review.
Start by naming what you actually need to verify
Before you open any database, write down the specific claim the paper supports in your work. A retraction matters differently depending on whether the paper is a load-bearing citation, a methods reference, or background context. Naming the claim tells you how much verification effort is proportionate and what you will need to change if the paper turns out to be retracted.
Three starting states are common:
- You are about to cite a paper and need to confirm it is still valid before it enters a manuscript.
- You are revising or closing a project and need to confirm that nothing in the reference list has been retracted since you last checked.
- You are starting a review and need a documented screening trail that shows how each included study was verified.
Each state needs the same core evidence, but the artifacts differ: a citation note, a revision log, or a screening record.
Build the verification record before you search
A retraction check is only auditable if it produces an artifact. Decide up front where the record will live and what fields it will contain. A minimal record has: the paper's full title, authors, journal, year, DOI or other identifier, the date you checked, each source you consulted, and what that source showed.
If the check belongs to a larger project, keep the record with the project rather than in a personal note. In WisPaper, Research Projects is the area for organizing work as a project, which gives the verification record a home alongside the rest of the materials. For a single paper you are about to cite, My Library lets you save or upload the paper and keep it organized with the sources you are screening.
Find the authoritative record for the paper
The first decision is which version of the paper you are checking. Preprints, conference versions, and journal versions can carry different status information. Identify the version of record and its persistent identifier before you look for retraction notices.
Cochrane's guidance on searching for studies emphasizes planning searches from explicit eligibility criteria and using multiple sources rather than relying on one. Apply the same logic here: treat any single source as one input, not as the verdict.
Practical steps:
- Record the exact title, author list, journal, and year as they appear on the version you are using.
- Capture the DOI or other persistent identifier.
- Note whether you are working from a preprint, a published version, or both.
If you need to locate the paper or confirm you have the right version, WisPaper search provides a scholarly search interface for finding academic papers. For a public explanation of the same capability, see WisPaper Scholar Search.
Check the publisher's page and the article's status markers
The publisher's page for the version of record is the place where a retraction notice is normally attached to the article. Open that page and look for status statements, linked notices, and any editorial notes attached to the article. Record what you see and the date.
This is also where you confirm that the identifier you recorded matches the article you are checking. A mismatch between the DOI and the title is itself a finding worth logging.
Cross-check bibliographic databases
Bibliographic databases index retraction notices and link them to the affected record. PubMed, for example, supports field tags, phrase searching, Advanced Search, and citation lookup, and its search details reveal how terms were translated and combined. That transparency is useful for an audit trail: you can record the exact query that produced your result.
A workable cross-check sequence:
- Search by the persistent identifier.
- Search by exact title in quotation marks.
- Search by author plus title.
- Open the full record and read the status and any linked comments or notices.
- Save or export the record so the check is reproducible.
Google Scholar's help documentation describes searching by author and title, date sorting and limits, Cited by and Related articles, citation export, and ways to locate alternative versions of a work. Those features help you find other versions and see how the paper has been cited, but the status of the version of record still comes from the publisher and the indexing database.
Use citation context as a secondary signal, not a verdict
Citation trails can tell you that a paper has been discussed, corrected, or superseded, and they can surface commentary you would otherwise miss. They are a secondary signal. A citation that mentions a correction is a lead to follow, not proof of retraction status.
When you follow a lead, return to the authoritative record and log what you found there. Keep the citation-context note separate from the status finding so a later reader can see which was which.
Screen the whole reference list, not just the paper in front of you
When you are revising a manuscript or closing a project, the task is usually broader than one paper. Cochrane's guidance recommends documenting the search process, deduplicating records, screening in stages, and linking multiple reports of one study. Those practices map directly onto a reference-list audit.
A staged screen looks like this:
- Stage 1: list every reference with its identifier and the date it was last verified.
- Stage 2: flag references that have never been verified or were verified before your project's cutoff.
- Stage 3: run the full check on flagged references and record the result for each.
- Stage 4: link multiple reports of the same study so one retraction does not leave duplicate entries unaddressed.
Cochrane also recommends rerunning searches before publication. For a reference audit, that means scheduling a final pass close to submission rather than relying on checks done months earlier.
Worked example: auditing a manuscript's reference list before resubmission
Consider a corresponding author preparing a revised manuscript. A coauthor has added new citations during revision, and the reference list has not been audited since the first submission.
Step 1 — Define the scope. The author lists all references and marks which ones support load-bearing claims. Those get priority.
Step 2 — Build the record. Each reference gets a row: title, authors, journal, year, identifier, last-verified date, sources checked, and result.
Step 3 — Check the version of record. For each priority reference, the author opens the publisher page and records the status shown.
Step 4 — Cross-check the database. The author searches by identifier and by exact title, opens the full record, and exports it. The exported record and the query used are saved with the audit row.
Step 5 — Screen the remainder in stages. Non-priority references are checked in a second pass, and duplicate reports of the same study are linked so a single finding covers all of them.
Step 6 — Rerun before submission. A final pass confirms that nothing changed between the audit and the submission date.
Step 7 — Hand off the trail. The completed audit table goes into the project record so a coauthor or successor can see exactly what was checked and when.
If the audit is part of a larger literature review, Scholar Agent is the area for longer research tasks that fit literature review, inspiration, or reproduction use cases, which is where a reference audit often sits.
Verify the citations themselves, not only their retraction status
A retraction check confirms that a paper exists and is still valid. It does not confirm that the citation in your manuscript points to the right paper. A reference can be unretracted and still be wrong: mismatched authors, wrong year, or a BibTeX entry that does not correspond to any real record.
Run a separate citation screen on the bibliography. TrueCite is built for BibTeX citation screening and checks BibTeX against real academic databases to flag hallucinated references. Pair that screen with the retraction audit so each reference has both a validity result and a citation-accuracy result.
Ask questions against the papers you have already collected
Once papers are saved in your library, questions about them can be answered from that collection rather than from a fresh search. Library QA supports asking questions based on papers uploaded to My Library or added from search results. Use it to check what a saved paper says about its own status or to compare how several saved papers describe a shared method, then confirm any status finding against the authoritative record.
Reusable checklist and completion criteria
Use this table as the artifact for each paper you check.
| Field | What to record |
|---|---|
| Title and authors | Exactly as on the version you are using |
| Journal and year | Version of record details |
| Identifier | DOI or other persistent identifier |
| Version checked | Preprint, published, or both |
| Publisher page | Status shown and date checked |
| Database record | Query used, record exported, date checked |
| Citation context | Any commentary found, kept separate from status |
| Citation accuracy | Result of the BibTeX or reference screen |
| Result | Clear status statement with date |
| Reviewer | Who ran the check |
Completion criteria. The check is ready to close when all of the following are true:
- The version of record has been identified and its identifier recorded.
- The publisher page has been checked and the status recorded with a date.
- At least one bibliographic database has been cross-checked, with the query and exported record saved.
- Any citation-context leads have been followed back to an authoritative record.
- The citation itself has been screened for accuracy.
- The result is stated plainly in the record, with the date and the reviewer named.
- For a manuscript or project, a final rerun is scheduled before submission or closure.
How to know the result is ready
A retraction check is ready when a reader who was not present for the work could repeat it from your record and reach the same status statement. That means the record names the version checked, the sources consulted, the queries used, and the date. It also means the result is stated as a status with a date attached, not as a general impression.
If any of those elements is missing, the check is incomplete regardless of how confident you feel about the answer.




