How to check whether a paper has been corrected or retracted

Before citing a paper, researchers should check whether it has been corrected, retracted, replaced, or otherwise flagged by the publisher or scholarly record. A citation can be real and still be risky if the paper no longer supports the claim you want to make.

This check is especially important when using AI tools. AI may suggest papers, summarize papers, or format references, but it may not warn you that a paper has a correction notice or retraction history.

This guide explains how to check a paper's publication status, what signals to look for, and how to decide whether a source is safe to cite.

A corrected paper has an official notice that changes or clarifies part of the published record. The correction may fix author details, figures, tables, methods text, errors in analysis, or wording that affects interpretation.

A correction does not always mean the paper is unusable. The key question is whether the correction affects the claim you plan to cite.

Ask:

• What was corrected?
• Does the correction affect results or conclusions?
• Is the corrected version available?
• Does your planned citation depend on the corrected section?
• Should you cite the correction notice as well?

Do not treat all corrections as equal. Some are minor metadata fixes. Others change how the paper should be read.

A retracted paper has been formally withdrawn from the reliable published record. Retractions may happen because of serious error, unreliable data, ethical issues, duplicate publication, peer review concerns, or misconduct.

A retracted paper should not usually be cited as evidence for a claim. It may be cited only when the retraction itself is the topic, such as a discussion of research integrity or publication history.

If you cite a retracted paper without realizing it, the problem is not just bibliographic. It can weaken the credibility of your literature review.

That is why status checking should happen before the final bibliography is locked.

These terms describe different publication-status signals.

• Retraction: the paper is withdrawn from the reliable record.
• Correction or erratum: part of the record is corrected.
• Corrigendum: a correction, often issued by authors or publisher depending on journal usage.
• Expression of concern: the publisher warns readers about a possible issue while investigation or clarification is ongoing.
• Withdrawal: a version, often early or online-first, is removed before or around formal publication.
• Replacement: a newer version may supersede an older record.

The exact terminology can vary by publisher. Always read the notice itself instead of relying only on the label.

Start with the publisher page or DOI landing page. The publisher usually has the official record, linked notices, updated versions, and visible banners.

Check:

• The article page.
• DOI landing page.
• Publisher notices.
• Linked correction or retraction notices.
• PDF watermark or header.
• Article history.
• Crossmark button if available.

Crossref's Crossmark is designed to help readers check whether they are viewing the current version of a scholarly document. When a publisher uses it, the Crossmark button can point to update information.

Publisher status is the first stop because it is closest to the official record.

Use the DOI as the stable identifier for the paper. Search the DOI in the publisher site, Crossref, academic databases, and retraction databases.

A DOI check helps because titles can change, preprints can have similar names, and AI-generated references may contain small metadata errors.

Check:

• Does the DOI resolve?
• Does it match the title and authors?
• Does the landing page show a notice?
• Are there linked updates?
• Does another record point to a correction or retraction?
• Is the paper a preprint, accepted manuscript, or final published article?

If the DOI does not match the citation, treat the reference as unresolved until you can verify it.

Citation databases and retraction databases can add another layer of checking. They may surface notices that are easy to miss on a publisher page.

Use them to look for:

• Retraction notices.
• Correction notices.
• Expressions of concern.
• Citation warnings.
• Links between the original paper and the notice.
• Publication-status metadata.

Do not rely on only one database. Metadata can lag or differ between systems. If a status issue matters to your argument, check the publisher record as well.

This same logic applies to AI tools. A tool can point you toward a concern, but the final citation decision should be checked against source records.

Decide by comparing the correction with the claim you want to support. A corrected paper may still be usable if the correction does not affect the cited claim.

Use this decision path:

1. Read the correction notice.
2. Identify the section affected.
3. Compare it with your planned citation claim.
4. Check whether the corrected version is available.
5. Decide whether to cite the paper, the correction, or another source.
6. Add a note in your citation record if the correction matters.

For example, a correction to an author affiliation may not affect your literature review synthesis. A correction to a results table may directly affect your claim.

The decision depends on the link between the notice and your use of the paper.

If a paper is retracted, do not use it as positive evidence for a substantive claim. If the paper is relevant because it was retracted, make that status explicit.

Options include:

• Exclude it from evidence synthesis.
• Mention it only as part of research-integrity context.
• Replace it with later reliable studies.
• Cite the retraction notice if discussing the retraction.
• Record the exclusion reason in the review notes.

If the review includes a formal screening process, a retracted paper should have a clear status and decision record.

For source-set documentation, see literature search log template for AI-assisted reviews.

Publication status is only one part of citation checking. A paper can be valid and still not support the sentence where it appears.

Check:

• Does the paper actually make the claim?
• Is the claim from results, discussion, background, or another cited paper?
• Is the finding direct or indirect?
• Is the population or method relevant?
• Does the claim overstate what the paper found?
• Has a correction changed the relevant section?

This step catches a common AI-assisted writing problem: real citations attached to unsupported statements.

For broader verification, see citation hallucination checkers for AI-generated references.

Turn the advice into a repeatable workflow by defining the decision you need to make, the evidence required for that decision, and the record that will prove how the decision was made. In citation verification, the problem is rarely one missing tool. The problem is usually that search, reading, checking, and writing happen in separate places without a shared rule.

Use a short operating routine:

• Name the review question or subquestion.
• Define the source set you are working from.
• Decide what counts as enough evidence for the next step.
• Apply the same criteria to every paper in that step.
• Mark uncertain cases instead of forcing a clean answer.
• Keep source locations for claims that may enter the final review.
• Review the workflow after each major search, screening, or writing session.

This routine keeps the work moving without making the review careless. It also gives supervisors, collaborators, and future you a way to understand why the source set changed.

Record the pieces that would be hard to reconstruct later. You do not need a diary of every click, but you do need enough detail to explain the path from question to source to claim.

For this topic, the most useful record usually includes metadata, DOI records, source status, claim support, and final citation use. Add the date, tool or source used, reviewer status, and next action. If AI assisted the step, write down what it helped with and what a human checked.

The record should distinguish discovery from evidence. A tool may help find a paper, but the paper itself must support the claim. A summary may help triage a source, but the original source should support any statement that appears in the literature review.

Before writing, check whether the workflow has produced usable evidence or only useful notes. Notes help you think. Evidence supports a sentence.

Ask:

• Which claim will this source support?
• Is the claim narrower than the evidence?
• Have methods, sample, outcome, or concept details been checked?
• Are limitations visible?
• Are conflicting papers handled rather than ignored?
• Is the citation real, current, and relevant?
• Can another reader understand how this source entered the review?

If the answer is unclear, keep the point in notes rather than moving it into the draft. This is the small pause that prevents AI-assisted research from becoming polished but weak writing.

WisPaper helps researchers find, organize, and inspect academic papers, while TrueCite provides a focused citation-verification step. TrueCite, powered by WisPaper, checks BibTeX files against real academic databases to flag hallucinated references.

That is useful when a researcher has AI-generated references, exported BibTeX, or a bibliography that needs existence and metadata checks before deeper source review. WisPaper's broader workspace can also help researchers save or upload papers into My Library and use Library QA to ask questions based on their own source set.

TrueCite should be treated as a citation verification aid, not as a substitute for reading the paper. Researchers still need to check whether each source supports the claim, whether notices affect the citation, and whether the source belongs in the review.

Check citations with TrueCite