How FreeCitation builds a citation

Every citation the tool gives you is built in the same order: find the real record for your source, then format it with the published rules of the style you picked. This page shows each step, where the data comes from and where you should still look twice.

Published 31 October 2026

Step 1: working out what you pasted

The tool first decides what kind of identifier you gave it. Anything that starts with 10. followed by a registrant code, or contains doi.org, is treated as a DOI. A 13 digit number starting with 978 or 979 is an ISBN. Anything starting with http is a web address. Everything else, such as a title or a description, is treated as free text.

Step 2: finding the real record

Journal articles and other DOIs. A DOI is resolved through Crossref, the registry that publishers such as Springer, Elsevier and Wiley deposit their records with. The authors, title, journal, volume, issue, page range and year all come from the publisher's own deposit, which is why DOI citations are the most reliable thing the tool makes.

Books. An ISBN is looked up in Open Library, the Internet Archive's open catalogue. Its records give the author, title, publisher and year. Some older or small press books have thin records there, so check the edition and the publisher against the copyright page.

Web pages. For a web address the tool downloads the page and reads the structured data the site publishes for search engines first. Most news sites and blogs mark up the headline, the author and the publication date there, and those beat anything else on the page. When that is missing, it falls back to the page title, the Open Graph tags and the standard author and date meta tags. The access date is the day you made the citation.

Step 3: formatting in your style

DOI and ISBN records are formatted with the Citation Style Language, an open standard in which each style is written as a file of rules. These are the same files behind Zotero and Mendeley, maintained in a public repository. This is the file used for each style:

Web pages do not have a shared record format, so the tool fills their details into a template written for each style: author, date, title, site name, address and, where the style asks for it, the access date.

When a style manual changes, the community files can lag behind it. The Chicago Manual of Style 18th edition, for example, dropped the place of publication for books in 2024. Our guide to the Chicago 18th edition changes lists the differences to look for.

Step 4: when AI steps in

If the tool finds no record at all, for example when you type only a title, it asks a Google Gemini model to draft the citation from what you typed, in the style you picked. That result is marked AI generated in the tool, with a reminder to verify it. Free visitors can use up to 10 of these AI lookups an hour.

An AI draft can be wrong. It can guess a date, mix up a page range or, with a vague title, describe a source that does not exist. Our article on fake AI citations explains how that happens and how to spot it. If you have a DOI or an ISBN, paste that instead of the title and the tool will use the real record.

What you should still check

• Author names: spelling, initials and the order of authors against the source itself.
• Capitalisation: APA uses sentence case for article titles, MLA and Chicago use title case. Records often arrive in the publisher's own casing.
• Dates and page ranges, especially for web pages and anything marked AI generated.
• Italics after you paste. The tool copies with formatting, but some editors strip it.
• Your instructor's own rules, which always win over the general style.

Questions

Found a citation the tool got wrong? Send the source and the style to [email protected]. How we write and correct the guides on this site is set out in our editorial policy.