URL to Citation

Well built pages describe themselves in machine readable ways, and those declarations are what a citation can be built from. News sites and journals usually publish all of it; personal blogs and government pages often publish almost none, which is why results vary so much between sites.

• Schema.org JSON-LD: author, datePublished, headline, publisher
• OpenGraph tags: og:title, og:site_name, article:published_time
• Standard meta tags: author, description, citation_title

The address you paste decides what can be read, and a surprising number of failures come from pasting the wrong form of it.

• Use the canonical article URL, not a search result or a redirect from a link shortener
• Strip tracking parameters such as utm_source, which sometimes route to a different page variant
• If the page is a PDF hosted on a website, upload the file instead so the DOI inside can be read
• For a paper on a publisher site, prefer its DOI: it is stable and gives far better metadata
• For a page behind a login, the fetch sees only the paywall, so enter the details manually

The URL is fetched server side and its structured metadata is parsed, with Schema.org preferred over OpenGraph and OpenGraph over plain meta tags. If the page turns out to be an article with a DOI, that DOI takes over and the reference is built from the registry record instead, which is always the better source. Whatever is found is shown to you, so you can see which fields came from the page and which are missing.

Web citation is the messiest part of any citation tool, and some of the gaps are in the page rather than in the tool.

• No author. Many organisational pages have none, and the styles have a rule for this: use the organisation as the author. APA and Harvard both do this, MLA starts the entry with the title.
• No date. Use n.d. in APA and Harvard. This is also the case where a retrieval date genuinely belongs in the reference.
• The page changes. Wikis, dashboards and live statistics pages need a retrieval date because the content you cite may not be there next month.
• JavaScript only pages. If the content is rendered in the browser rather than sent by the server, there is often nothing in the HTML to read.
• Social media posts. These need the platform's own conventions, including the handle and the full text of short posts, and are usually faster to enter by hand.

Web references diverge more between styles than any other source type. Watch the URL in particular: MLA drops the protocol, Harvard introduces it with Available at, and only some styles ask for an access date.

This page does not lock the output format. Find the source first, then switch between styles, or start from a style page if your department only ever uses one.