Methodology
A citation is only useful if it is right. This page states exactly what we check against, what we fetch, and what we deliberately do not do.
Nothing here is written by a model
Formatting is deterministic: a set of rules per style, applied to structured metadata. A language model can produce a plausible volume number for an article that has none, and that single invented field is enough to fail a submission. If a field is missing upstream, we leave it empty and ask you for it.
Where the metadata comes from
- Crossref Journal articles, conference papers, books and chapters that carry a DOI. www.crossref.org
- Open Library Books by ISBN: title, authors, publisher, place and year. openlibrary.org
- PubMed Biomedical and life-science articles, including those without a DOI. pubmed.ncbi.nlm.nih.gov
Every generated reference names its source. Web pages are the exception: there is no key-less public service that reads a page's metadata from your browser, so for a URL we prefill what the address itself tells us and you supply the title and author.
What happens when we are not sure
A DOI, an ISBN or a PubMed ID is exact: there is one record and we use it. A title typed from memory is not. Catalogue search will always return something, so a free-text search here has to clear a relevance score and a word-overlap check before it is shown at all, and even then we list the candidates and let you choose. Nothing is selected for you, and a query that matches nothing says so instead of offering the closest thing it found.
That matters more than it sounds. A plausible reference to a paper you have never read, labelled as verified, is worse than no answer: you would have no reason to check it.
Capitalisation
Publishers register titles in their own house style, so the same article arrives in sentence case from one journal and headline case from another. Styles disagree too. We convert to whatever the chosen style requires, protecting acronyms, hyphenated compounds and anything carrying a digit, so DNA and COVID-19 survive the conversion.
Which edition of each manual we follow
| Style | Checked against | Last reviewed |
|---|---|---|
| IEEE | IEEE Reference Guide | August 2026 |
| ACS | The ACS Guide to Scholarly Communication, 3rd ed. | August 2026 |
| Turabian | Turabian, A Manual for Writers, 9th ed. | August 2026 |
| AMA | AMA Manual of Style, 11th ed. | August 2026 |
| ASA | ASA Style Guide, 6th ed. | August 2026 |
| Chicago | The Chicago Manual of Style, 18th ed. | August 2026 |
| Vancouver | ICMJE Recommendations and NLM Citing Medicine | August 2026 |
| BibTeX | BibTeX format and the biblatex manual | August 2026 |
| APSA | APSA Style Manual for Political Science | August 2026 |
| CSE | Scientific Style and Format, 9th ed. | August 2026 |
| NLM | Citing Medicine, the NLM Style Guide | August 2026 |
| APA | Publication Manual of the APA, 7th ed. | August 2026 |
| MLA | MLA Handbook, 9th ed. | August 2026 |
| Bluebook | The Bluebook, 21st ed. | Not yet published |
The style we have not built
Legal citation is the one style we have not published. Bluebook describes cases, statutes and courts rather than authors, titles and publishers, so it needs its own data model instead of a variation on the one behind every other style here. We would rather ship it late than ship it wrong.
What we do not store
Your bibliography lives in your browser's local storage. There is no account, no server-side copy, and no analytics attached to the sources you look up. Clearing your browser data clears your list, so export it first if it matters.
Corrections
If a reference comes out wrong, that is a bug and we want it. Send the input you used and the output you expected to hello@citeanystyle.com.