Methodology
How we verify a citation
Hallucite compares cited identifiers and publication details with available records. Decisions use deterministic, versioned rules; the result shows the evidence and limits of each check.
Why check references
Incorrect identifiers and publication details can make a reference difficult to locate. Whether you are reviewing someone else’s work or your own, check the bibliography and review the evidence for entries that need attention.
Verification sources
Sources are used when applicable to a reference. Coverage and availability vary; the result shows the evidence obtained.
- Crossref
Publication metadata and available retraction notices.
Data also held in our local mirror.
- PubMed
Biomedical publication metadata and available publication notices.
Data also held in our local mirror.
- NLM Catalog
Biomedical books and reports through the biomedical catalogue lookup.
- OpenAlex
Publication metadata and identifiers.
Data also held in our local mirror.
- CiNii
Japanese publication and book catalogue records.
- DBLP
Computer science publication records.
- Open Library
Book records through ISBN lookups and title searches.
- Google Books
Book records through ISBN and title searches.
- Semantic Scholar
Scholarly publication records and arXiv identifier lookups.
- DOI registry
DOI resolution and identifier checks.
A local mirror and provider lookups supply publication evidence. Sources are selected according to the reference; no complete coverage is promised. Missing entries, incomplete metadata and source outages can leave a reference unresolved.
Reading the results
Publication match — available evidence supports a match to a publication. Read the cited fields alongside the matched record; a match does not establish claim support.
Metadata discrepancy — one or more cited details differ from the publication record. Inspect the year, authors, title, venue and identifier before deciding what to correct.
Unresolved reference — the completed checks did not establish the cited work. Coverage gaps, incomplete metadata or an unavailable source may explain the result. Investigate further; this is not proof of fabrication.
Read each result label with its details. Parsing failures and source outages describe checking limits. Publication notices are reported separately where available; inspect their scope. No result establishes authorship, intent or whether a source supports a particular claim.
The errors we localize
When something is off, we point to the specific field rather than just raising a flag:
- Identifier mismatch — a DOI or PMID resolves to a publication whose details differ from the citation.
- Author differences — the cited author list differs from the publication record.
- Title differences — the cited title differs from the matched publication record or remains unresolved.
- Year, journal & cross-reference consistency — metadata that contradicts the real record.
Compare identifiers and metadata
A resolving identifier alone does not establish a correct citation. Hallucite compares the returned publication record with the cited title, authors and publication details. Differences are shown for review.
How we measure and what remains uncertain
Hallucite makes deterministic, versioned verification decisions and shows the evidence for each flag. Hosted in the EU.
Coverage gaps, incomplete metadata and source outages can leave a reference unresolved. Review flagged entries: a missing match does not prove fabrication, and a verdict does not establish authorship or whether a source supports a claim.
Benchmark results are still to come and will be published after review.
Read the methodology