Registry Coverage
StrictCite checks every reference against thirteen public bibliographic registries, queried directly from your browser. This page lists all thirteen, what each one holds, and which literature none of them reach.
A verdict of “no record found” means no record in these thirteen. It does not mean the work does not exist. Scholarship that predates the DOI, was published in a national-language journal, or appeared as grey literature is frequently absent from all of them at once — which is why an unmatched reference is marked UNVERIFIED, never REJECTED.
The reverse also holds: an answer is not the same as a fact. Crossref and DataCite both occasionally carry corrupted or duplicate records — a paper deposited more than once under fabricated metadata, an author list a publisher itself uploaded broken. StrictCite shows you which registry made a claim rather than asserting it as settled, and nothing is applied on your behalf unless two or more independent registries agree; a claim resting on a single source is left for you to weigh, not applied for you.
Identifier authorities
asked firstThe agency that issued an identifier is the authority on what it points to. When a citation carries a DOI or an ISBN, these are consulted before anything else, and their answer outranks an index that merely copied the record.
The DOI registration agency used by most scholarly publishers: journal articles, conference papers, books, chapters and preprints, deposited by the publishers themselves.
The DOI agency for datasets, software, theses, reports and preprint servers. It holds arXiv’s catalogue under arXiv’s own DOIs.
The resolver itself. It answers one question no index can: does this DOI resolve at all, and to what? A cited DOI that fails here is fabricated or mistyped.
The Internet Archive’s book catalogue: monographs, editions and imprints. Consulted for references that look like books rather than articles.
Indexes and specialists
second opinionsBroad catalogues and field-specific archives. They supply corroboration where an authority already answered, and coverage where none did — a preprint, a computer-science proceeding, an open-access article whose publisher deposits little.
An open index of scholarly works across every discipline, successor to Microsoft Academic Graph. The broadest single catalogue on this list.
The Allen Institute’s corpus, drawn from publisher feeds and web-scale harvesting. Often holds records the registration agencies missed.
The US National Library of Medicine’s biomedical index, and the authority behind a cited PMID.
The European life-sciences counterpart, with wider preprint and full-text coverage than PubMed alone.
The computer-science bibliography. Conference proceedings are the field’s primary literature and are indexed here more reliably than anywhere else.
The Directory of Open Access Journals: articles in vetted open-access titles, including many outside the major publishers’ deposits.
The biology and health-sciences preprint servers, including the version history that tells you whether a preprint was later published.
CERN’s general-purpose repository: datasets, software releases, reports and conference material that never reached a journal.
A map of legal open-access copies for a given DOI. It is not a catalogue in its own right, but it confirms that a DOI corresponds to a real, retrievable article and returns the journal and year it was published under.
What those thirteen reach
Thirteen is a small number for a coverage claim, and it is the honest one. Registries are not publishers: a registration agency holds the records its members deposit, so querying Crossref reaches every publisher that registers DOIs there. The list below is not a partnership or an integration — it is simply where those records already live.
Elsevier and ScienceDirect, Springer Nature, Wiley, Taylor & Francis, SAGE, IEEE Xplore, the ACM Digital Library, IOP Publishing, the American Physical Society, and the university presses — each deposits its own DOIs with Crossref, which is why a reference to any of them can be checked against the publisher’s own record.
arXiv, Figshare, Dryad, the Open Science Framework and Harvard Dataverse register their DOIs with DataCite. arXiv’s own API sends no cross-origin header and cannot be called from a browser at all; the DataCite record carries the same metadata.
Computer science publishes in conferences rather than journals, and those proceedings are indexed in DBLP more completely than anywhere else — NeurIPS, ICML, CVPR, ICCV, ACL, EMNLP and CHI among them, including years whose DOIs were never registered.
Crossref and DataCite are two of roughly a dozen DOI registration agencies. The others — JaLC in Japan, mEDRA in Europe, KISTI in Korea, and the rest — are reached through the resolver itself, which answers for every DOI ever issued regardless of who issued it. A DOI outside our two agencies is still checked for existence, even where full metadata is thinner.
Both harvest institutional repositories and national archives beyond the deposit chain, which is where a work sits when its publisher registers nothing. Coverage here is genuinely uneven, and it is the reason both are consulted rather than either alone.
The publishers and archives named above are named to describe where records come from. StrictCite is not affiliated with, endorsed by, or partnered with any of them.
Known fraudulent registrations
A DOI that resolves and agrees with every registry on every field is not automatically a real citation to a real record — it can be a duplicate registered purely to look like one. We found five such duplicates of “Attention Is All You Need” sitting on Crossref under a fabricated 2025 date, byte-identical to the real paper’s metadata in every field this engine compares. No amount of cross-checking against Crossref, DataCite, or anything that proxies either one catches that on its own, because the fraud is the registration itself, not a discrepancy in what it says.
A short, hand-maintained list of DOI prefixes already confirmed to be citation-spam registrations rather than publishers. A citation whose DOI matches one is flagged regardless of what every other field says — overriding a clean match, deliberately, since a spam duplicate is built to produce one. It is a blocklist, not a fraud detector: it catches a scheme once someone has found and confirmed it, and nothing before that. The first entry, and how we found it, is written up here.
What none of these cover
Stated plainly, because a verification tool that hides its blind spots is worse than none. If your bibliography is largely made of the following, expect unmatched references that are perfectly real:
- • Books and chapters published before ISBN registration was routine, and many university-press monographs since.
- • Journals published in languages other than English, particularly those outside the major indexing programmes.
- • Grey literature: working papers, government and agency reports, standards, theses held only by a single institution, and internal technical memoranda.
- • Conference proceedings with no assigned DOI, outside computer science.
- • Archival and primary sources: manuscripts, legal instruments, newspapers, correspondence.
- • Scopus and Web of Science, which require an institutional licence and publish no open interface. StrictCite does not query them, and no tool can do so from your browser.
For these, StrictCite still checks internal consistency — whether a DOI resolves, whether page ranges and years are coherent, whether an identifier belongs to the work it is attached to — and reports the reference as unverified rather than guessing.
Why there are no record counts on this page
Every registry above grows weekly, and a figure quoted once becomes wrong quietly. Rather than print a number we would have to keep true, each entry describes what the registry holds. If you need current totals, each publishes its own.
Check your bibliography against all thirteen
Your manuscript never uploads. Only the identifiers and titles of the works you cite are ever sent.