Citation Generator FAQ: How the Tool Works, What It Supports, and How Accurate It Is
This page collects the questions readers ask most often about how the citation generator on this site works — which styles it supports, what to do when a citation looks wrong, what the saved-references feature does, and the basics of how the underlying engine builds references. For the details on each style’s formatting rules, follow the links to the dedicated style guides; this page is about the tool itself.
The shortest answer: the tool produces citations in seven styles using the official CSL definitions, supports URLs, DOIs, ISBNs, and manual entry, saves your references locally in your browser, and is free. If a citation looks wrong, it is usually because the source did not expose complete metadata — fix the missing fields in the generated entry and verify against the original source.
How the tool builds citations
Every reference is rendered using the Citation Style Language (CSL) project — the same machine-readable style definitions used by Zotero, Mendeley, and Pandoc. This site ships the current CSL definition for each supported style and runs them through citeproc-js, the reference implementation. The practical consequence is that a citation produced here follows the same formatting rules as any CSL-compliant tool, which is the standard for academic citation tooling.
When you paste a URL, the extractor fetches the page and reads multiple signal sources in parallel — JSON-LD, Open Graph, Twitter cards, microdata, meta tags, and HTML heuristics — and merges them into one best-confidence reference. The architecture is multi-signal rather than single-source because no single metadata channel is reliable across the open web: news sites use one approach, academic journals another, blog platforms a third. Pulling from several signals at once and reconciling them produces a more accurate citation than reading any single source alone.
For books, an ISBN is resolved against Google Books and Open Library. For journal articles, a DOI is resolved against Crossref and OpenAlex. Both pipelines fall back to manual fields when the external services do not return complete data.
What the tool supports
| Source type | Input | What the extractor uses |
|---|---|---|
| Website / web article | URL | JSON-LD, Open Graph, Twitter cards, microdata, meta tags, HTML heuristics |
| Journal article | DOI | Crossref, OpenAlex |
| Book | ISBN | Google Books, Open Library |
| Other source types | Manual entry | Form fields filled by hand |
The supported automatic source types are the ones with reliable external metadata. Source types that do not have a single canonical identifier — personal interviews you conducted, in-class lectures, archived correspondence, unpublished manuscripts — are handled through the manual-entry form. The per-source-type how-to guides explain what fields to fill in for each case.
How accurate the citations are
Output that comes from a complete set of source metadata is accurate to the style manual. A DOI for a journal article, an ISBN for a book, a well-formed JSON-LD record on a web page — these produce citations that match what the style manual specifies.
Output from incomplete metadata is only as accurate as the metadata available. A web page that does not expose its author, its publication date, or a canonical URL produces a partial citation. The generator fills the gaps it can identify (using the URL as a partial publication date proxy, the domain as a partial publisher hint) and flags the rest for you to fill in. The shortest reliable rule for any submission: verify the generated entry against the source itself before submission, especially for any source the extractor could not fully extract.
If you find an entry where the generator’s output disagrees with the official style manual — not just incomplete because of source metadata, but actively wrong — that is a bug. The About page lists how to report it; corrections are treated as priority work.
What to do if a citation looks wrong
Three cases account for almost every “this citation looks wrong” message:
Missing fields. The source did not expose enough metadata for the generator to fill every required slot. Fix: open the generated entry and fill the missing fields manually, then re-render.
Wrong style edition assumed by the engine. The CSL definition for a given style sometimes lags behind a new edition by a few months. If you are writing for the latest edition and the output reflects an older convention, the dedicated style guide covers what the current edition specifies; reconcile by hand.
Style convention disagreement. A few engine outputs follow the style’s CSL definition but differ slightly from convention in printed style manuals (Cite Them Right Harvard’s exact punctuation, NLM Citing Medicine’s date format). The shipped style guides on this site reflect what the engine produces; if your assignment requires the printed-manual variant exactly, adjust the output by hand. The differences are small but real.
In all three cases the same principle applies: a citation is only useful if it matches what the source-of-truth style manual says. Verify before submission.
What “My References” does
The My References page stores every citation you generate, in your browser’s local storage. There is no account, no sign-up, and no server-side reference storage. When you generate a citation, it is saved locally and visible on the references page; when you return to the site, the saved references are still there as long as you have not cleared your browser data.
The feature exists so a research project that takes weeks does not require re-entering source URLs every session. Each saved reference includes the source data, the rendered citation in your selected style, and a timestamp. You can switch the style on the references page to re-render all saved entries in a different style.
What the feature does not do: sync across devices, back up to a server, export to Zotero or Mendeley. An export feature is on the roadmap; for now, treat the references page as a working draft list rather than a permanent archive.
Privacy
There is no account, no email signup, and no server-side storage of your reference list — references live in your browser’s local storage. Standard web logging (IP address, browser type, referring URL) applies, and we use privacy-friendly, cookieless Cloudflare analytics; the site does not currently serve ads or use advertising cookies. The full details — what is collected, how long it is retained, and your rights — are documented on the privacy page.
Browser and device support
The tool runs in any current browser — Chrome, Edge, Firefox, Safari — on desktop and mobile. The citation engine runs server-side on Cloudflare Pages Functions, so your browser never has to download a heavy formatting library — only the small amount of JavaScript needed to submit a source and display the result. Mobile layouts are responsive; the main use case is reading the generated citation to copy it into a word processor on the same device or a different one.
The site does not require JavaScript to read the guides (every guide page is statically rendered); JavaScript is required for the citation generator and the references page, which manage state in the browser.