A paper that cites a retracted study inherits its problem: the citation props up a claim with evidence the scientific record has formally withdrawn. Nothing breaks automatically, no alert reaches the author, and the reference can sit unnoticed for years, quietly passing the retracted finding along to every reader who follows it.
The timing trap: valid when you cited it, retracted before you submit
A retracted reference is usually a timing problem rather than a diligence failure. Retraction always happens after publication, often long after, so when you read a paper, judge it sound, and cite it, you are citing the record as it stands that day. Between that day and the day your manuscript reaches an editor sit months of drafting, co-author rounds, and revisions, and a retraction can land anywhere in that window. Nothing notifies you when it does. Publishers post retraction notices on their own sites, but no system pushes those notices to the authors who cited the work, and reference managers do not reliably flag them either. The citation that was defensible when you wrote it becomes a reference to withdrawn evidence, and it looks no different from any other entry in your bibliography.
The window extends past submission. Peer review adds months, so a reference can be retracted while your paper is under review, and a bibliography that was clean at submission may no longer be clean at acceptance. Screening once, at the moment of writing, is therefore never enough: the retraction record is a moving target, and the check belongs at the last responsible moment, ideally at submission and again before final acceptance. None of this reflects carelessness. It reflects a structural gap between a static reference list and a scientific record that keeps correcting itself, which is why the trap catches even careful researchers.
How common are retractions now?
Retractions are at record volume, which means the odds that any given bibliography touches one keep worsening. More than 10,000 research papers were retracted in 2023, a record annual figure, as reported by Nature. Much of the recent growth comes from bulk cleanups: journals withdrawing large batches of compromised papers at once, often after paper-mill activity or manipulated peer review comes to light. Retraction is no longer a rare correction of a single flawed study; it is a routine, high-volume maintenance process for the scientific record.
For a citing author, the reasons behind a retraction matter less than the fact of it. Whether the underlying paper fell to honest error, image manipulation, or fabricated data, the evidential support it lent your claim is gone, and your citation now points at withdrawn work. The volume trend makes this everyone's problem. Reference lists commonly run to dozens of entries, drawn from fast-moving fields where scrutiny is intense, and every year of record-setting withdrawals raises the chance that one of those entries has quietly changed status since you added it. A bibliography assembled over two or three years of a project ages in ways its author never sees, and the record pace of withdrawals documented by Nature means it ages faster now than it ever has.
What does citing retracted work cost?
Citing retracted work costs credibility first and accuracy second, and the bill arrives at several desks:
- With reviewers and editors. A reviewer who recognizes a withdrawn study in your reference list starts questioning the care behind everything else in the manuscript. At best the discovery prompts a revision request; at worst it colors the entire review.
- In the published record. Studies repeatedly find that retracted papers keep accumulating citations after retraction, often with no acknowledgment that the work was withdrawn. Each uncritical citation launders the withdrawn finding back into the literature, where new work builds on it.
- Downstream of the literature. In clinical and translational fields, conclusions resting on withdrawn evidence can shape systematic reviews, guidelines, and practice long after the original paper was flagged.
For journals the cost compounds. A published paper that leans on a retracted source draws reader complaints, correction workflows, and editorial time, which is why reference screening has been moving from a courtesy into routine diligence for editorial offices. For authors the asymmetry is stark: verifying a bibliography takes minutes, while a single retracted reference discovered by a reviewer, or by a reader after publication, can cost a correction notice attached to your name and a dent in the trust your next submission depends on.
How do you screen a bibliography for retractions?
Screening a bibliography for retractions became practical at scale in September 2023, when Crossref acquired the Retraction Watch database and made it openly and freely available. Before that, retraction notices were scattered across publisher sites, and checking a long reference list meant searching entries one at a time. Crossref calls the combined dataset the largest single open-source database of retractions, and because it is open, a submission workflow can now match every DOI in a bibliography against it in a single pass (Crossref).
Screening now has a second job: confirming that the references exist at all. A peer-reviewed study in Scientific Reports found that a majority of the bibliographic citations generated by GPT-3.5 were fabricated, and that GPT-4 still produced a substantial share of fabricated or erroneous ones. Even citations that pointed to real works frequently contained substantive errors, the same Scientific Reports testing found. As AI-assisted drafting spreads, a reference list can fail in three distinct ways, and one screening pass should look for all of them alongside plagiarism and the rest of what a full manuscript check covers:
| Failure mode | How it enters a bibliography | What screening looks for |
|---|---|---|
| Retracted reference | The source was valid when cited and withdrawn later, before or after submission | A match between the reference and an entry in the retraction record |
| Nonexistent reference | AI-assisted drafting produced a plausible citation to a work that was never published | A reference that fails to resolve to any indexed publication |
| Erroneous reference | The work exists, but the authors, year, journal, or identifier are wrong | Metadata that disagrees with the resolved record |
The common thread is verification against the live record rather than trust in the reference list as written. A bibliography is a set of claims about the literature, and like any claims, they can be checked. The same pass catches the newer failure mode: references that never existed at all, which is how hallucinated citations enter a manuscript.
How Octym helps
Octym checks the references in a manuscript against the retraction record as part of its review, and those reference checks are backed by the Retraction Watch database. A reference that matches a retraction, fails to resolve, or carries inconsistent metadata is surfaced as a signal with the supporting evidence, traced to its source. Octym never issues a verdict; a person decides whether to replace, keep, or contextualize the citation. That turns retraction screening from a post-publication surprise into a routine part of preparing a submission.