The Retraction Watch Database is the largest structured index of retracted, corrected, and concern-flagged scholarly articles in existence — over 60,000 records as of 2026 — and it is free to search with no registration required, at retractiondatabase.org. This guide covers how to actually use it: which fields you can search, how to combine and filter a query, what the reason-category taxonomy means when you’re reading results, and the practical workflows — due-diligence checks on a candidate, hire, or collaborator; literature-review hygiene before you cite something; and misconduct-pattern research — that researchers and research administrators actually use it for.
If you’re looking for what Retraction Watch is and who runs it, see the Retraction Watch dictionary entry. This guide assumes that context and focuses entirely on using the database itself.
Accessing the database
The database lives at retractiondatabase.org and is operated by the Center for Scientific Integrity (CSI), the US 501(c)(3) nonprofit that also runs the Retraction Watch blog. It is organizationally and financially independent of CASRAI and of any publisher. As of an October 2024 interface change, the search form requires at least one field to contain a value before it will run a query — you can no longer submit a fully blank search to browse the entire dataset. The same underlying data is also distributed through Crossref’s REST API under a September 2023 data-sharing agreement, which is the better route if you need to check retraction status programmatically (e.g. against a reference list or a grant application’s bibliography) rather than one record at a time in a browser.
What fields you can actually search
The search form exposes several fields, and understanding which are free-text and which are constrained dropdowns changes how you should query them:
- Author — free-text name search. Useful for checking a specific individual’s publication record, but be aware of name-collision risk with common names; cross-check hits against affiliation or subject before concluding a match is the person you mean.
- Title — free-text search of the retracted article’s title.
- Journal — offered as a dropdown of known journal titles, so you get exact matches rather than fuzzy text matching. Useful for auditing a specific journal’s retraction history, e.g. before submitting to it or evaluating it as a venue.
- Publisher — also dropdown-based; useful for a portfolio-level view across a publisher’s whole journal list (relevant, for example, when investigating a publisher-wide pattern the way Hindawi’s mass retractions surfaced after 2022).
- Affiliation(s) — a free-text field matching institution name or any part of a listed address. This is the field to use for an institution-level due-diligence check.
- Country(ies) — narrows by the country associated with the record.
- Article Type(s) — filters by publication type (e.g. journal article, conference proceeding, book chapter).
- Original Paper Date and Retraction/Notice Date — both accept date ranges, so you can separate “when was it published” from “when was it retracted” — a genuinely useful distinction, since the gap between the two is itself a signal (a paper retracted within a year of publication reads very differently from one that stood unchallenged for a decade).
- PMID or DOI — the fastest path when you already have a specific article identifier and just need a yes/no retraction check.
Search mechanics: Boolean logic, wildcards, and combining filters
Once you’re searching more than one field, the database applies Boolean AND logic across fields by default — an author search plus a journal search returns only records matching both. Within a single free-text field, you can use AND, OR, and AND NOT operators explicitly, and wildcard characters (*) can be placed before or after a term to broaden matching (useful for name variants or partial institution names). Where a field is a dropdown, you can select multiple values; multiple dropdown selections default to OR logic but can be switched to AND or AND NOT. Practically, this means a query like Affiliation containing a university name, combined with a Retraction Date range of the last five years, and a Reason category limited to fabrication/falsification, is a real, executable query — not something you have to reconstruct by exporting and filtering results yourself.
Interpreting the Reason categories
Every record in the database carries one or more Reason tags, assigned by CSI’s own team from the retraction notice’s stated language plus, where available, information from the underlying investigation. This taxonomy (documented in the database’s own Appendix B) is granular — dozens of distinct tags — and reading it correctly matters, because lumping every retraction into “misconduct” is a common and misleading oversimplification. The categories fall into rough groups:
- Misconduct — Falsification/Fabrication of Data, Image, or Results; Plagiarism of Text/Data/Image; Duplication; Salami Slicing; Paper Mill; False/Forged Authorship or Affiliation; Manipulation of Data/Images/Results; Misconduct by Author/Company/Institution/Third Party.
- Honest error — Error in Data/Methods/Results/Conclusions; Error in Image/Text; Error in Materials/Cell Lines; Error in Analyses; Error by Journal/Publisher/Third Party.
- Research ethics and approval gaps — Lack of IRB/IACUC Approval; Informed/Patient Consent None or Withdrawn; Ethical Violations; Concerns about Human/Animal Subject Welfare.
- Publication-process issues — Compromised Peer Review; Concerns with Peer Review; Rogue Editor; Conflict of Interest; Breach of Policy by Author.
- Data and content concerns — Original Data not Provided/Available; Unreliable Data/Image/Results; Results Not Reproducible; Contamination of Cell Lines/Tissues or Materials.
- Legal and administrative — Civil or Criminal Proceedings; Legal Reasons and/or Threats; Copyright Claims; Notice — Lack of or Limited Information; Retract and Replace.
- Newer/AI-related categories — Computer-Aided/Generated Content (text generators, randomizing algorithms, generative AI) and Hoax Paper (a manuscript intentionally drafted with fabricated data to test a journal’s screening) — both added as the database’s taxonomy has kept pace with how retractions are actually happening.
The practical takeaway: always read the specific Reason tag(s) on a record rather than treating “retracted” as a single undifferentiated red flag. An Author Unresponsive or Lack of Approval retraction (a procedural/consent gap, often with no data-integrity implication at all) reads very differently from a Falsification of Data retraction, even though both appear identically as “retracted” in a citation manager or a Crossref lookup. For the mechanics of what a specific retraction notice itself is required to say, see Retraction Statement and NISO CREC, which standardizes the metadata and display conventions publishers use when they issue one.
Why researchers and administrators actually use it
Due-diligence checks
Search a prospective hire, collaborator, grant applicant, or guest editor by Author and cross-check any hits against Affiliation and date range before drawing a conclusion. This is standard practice in hiring committees, tenure and promotion review, editorial-board vetting, and pre-award grant screening — not an accusation in itself, since a retraction on a record can range from an honest correction gone through the formal retraction process to a serious misconduct finding, which is exactly why reading the Reason tag (not just counting hits) matters here.
Literature-review and citation hygiene
Before finalizing a manuscript, systematic review, or grant bibliography, checking key cited works against the database (or the equivalent Crossref API lookup, which is the more practical route for a long reference list) catches the specific failure mode of unknowingly building an argument on a retracted source — the Retraction Watch dictionary entry covers the Zotero integration that automates this check against a reference manager library. This connects directly to the discipline covered in CASRAI’s guide to writing a literature review and PRISMA and systematic review methodology guide — a systematic review’s inclusion/exclusion screening should routinely check retraction status, not just relevance.
Misconduct-pattern and portfolio research
Filtering by Journal, Publisher, Affiliation, or Country over a date range lets a researcher or integrity office look for patterns rather than single incidents — a spike in Paper Mill or Compromised Peer Review tags concentrated in a journal’s guest-edited special issues, or a cluster of retractions tied to one institution or lab, is the kind of signal that individual case-by-case lookups miss. This is the same investigative approach behind CASRAI’s guides on paper mills and tortured-phrases red flags and how a retraction actually happens, and complements the editorial-decision framework in COPE Guidelines Explained.
Limitations to keep in mind
The database is a secondary, editorially curated index, not the retraction notice itself — always follow through to the publisher’s actual notice for anything load-bearing (a hiring decision, a formal allegation, a manuscript rejection), since notice quality varies and CSI’s own Reason tagging is necessarily an interpretation of what the notice says. Coverage, while extensive since the 2023 Crossref partnership substantially closed the gap between the two datasets, is not guaranteed to be fully complete or perfectly current for every journal and region. And a retraction record, on its own, is not proof of individual wrongdoing by every listed author — multi-author retractions frequently involve co-authors who had no role in the underlying problem, which is exactly why COPE’s own retraction guidance calls for distinguishing culpable from non-culpable co-authors rather than treating an entire author list as equally implicated.
Frequently asked questions
Is the Retraction Watch Database free to use?
Yes. Searching retractiondatabase.org requires no registration, subscription, or fee. The underlying dataset is also freely available through Crossref’s REST API following the September 2023 Crossref–CSI data-sharing agreement.
Can I search by retraction reason directly?
Yes — Reason is a filterable field, drawn from the taxonomy described above (Appendix B in the database’s own user guide). You can, for example, limit a search to only Paper Mill or only Falsification of Data records within a journal, publisher, or date range.
What’s the difference between this guide and CASRAI’s Retraction Watch dictionary entry?
The dictionary entry defines what Retraction Watch is, who runs it, and its Crossref integration. This guide is the practical how-to: which fields and operators the search interface actually supports, how to read the Reason taxonomy, and the concrete workflows (due diligence, literature-review hygiene, pattern research) researchers and administrators use it for.
Does a retraction in the database always mean misconduct?
No. Reason tags span a wide range, from honest Error categories and procedural issues like Author Unresponsive, through research-ethics gaps, to explicit Misconduct and Paper Mill categories. Read the specific tag and, where it matters for a real decision, the underlying retraction notice, rather than treating every record as equivalent.
How current is the data?
CSI updates the database on an ongoing basis as new retraction notices are published, and the Crossref partnership means retraction status also now surfaces directly on a DOI’s Crossref /works record. As with any external, independently maintained dataset, treat a database check as a strong signal rather than a real-time guarantee for a record published in the last few days.
