How to Cite a Website Correctly

A website citation carries five pieces of information: who wrote the page, what the page is called, what site it sits on, when it was published, and where it lives. Put those five into your reference list in the order your style asks for, and the citation is correct. Nearly every website citation that comes back flagged is missing one of them, usually the author or the date, and the gap has been papered over with the site's homepage address.

Before you write anything down, run one check that saves a lot of rework. Is the thing you are citing actually a web page? A journal article in PDF form, sitting on a university server, is still a journal article. It gets volume, issue and page numbers, not a website citation. Cite a page as a web page only when the web page itself is the source: a news article, a government guidance note, a company's methodology page, a blog post, a dataset description.

The five slots, and what to do when one is empty

Here is the same source in the three styles students use most. The source is a fictional example, but the shape is real.

APA 7 Kowalska, A. (2024, March 12). How reading habits changed after 2020. Reading Institute. https://example.org/reports/reading-habits-2024

MLA 9 Kowalska, Anna. "How Reading Habits Changed After 2020." Reading Institute, 12 Mar. 2024, example.org/reports/reading-habits-2024.

Chicago (notes) Anna Kowalska, "How Reading Habits Changed after 2020," Reading Institute, March 12, 2024, https://example.org/reports/reading-habits-2024.

Now the empty slots, which is where most of the errors live.

No named author. The organisation is the author. In APA the reference starts with the organisation name and the site name is then dropped, so you get Reading Institute. (2024, March 12). How reading habits changed after 2020. https://…. Do not write "Anonymous", and do not shift the page title into the author position unless your style tells you to.

No publication date. APA uses (n.d.) and adds a retrieval date for pages designed to change: Retrieved March 12, 2024, from https://…. MLA puts the access date at the end: Accessed 12 Mar. 2024. A retrieval date is not decoration. It is your record of what the page said on the day you read it, and it only carries weight if you note it on the day you actually read the page.

No page title, only a heading. Use the heading that names that specific page, not the site's tagline. If the browser tab and the on-page heading disagree, take the on-page heading.

A long tracking URL. Strip everything from ?utm_source= onwards, and anything after # unless the fragment is genuinely part of the address. Then open the shortened link in a private window to confirm it still resolves.

Prefer a stable identifier over a raw URL

If the page offers a DOI, cite the DOI. Persistent identifiers exist precisely because ordinary web addresses do not last. Žero (2023) describes ISBN, ISSN and DOI as identifiers that are machine- and human-readable and that permanently identify and retrieve an object, with DOI covering articles, book chapters, general documents and datasets (s. 73). A DOI survives a site redesign, a domain change, and a publisher migration. A URL survives none of those reliably.

The practical order is: DOI first, then a stable publisher or repository link, then the plain URL as a last resort.

Why the date on a web citation does real work

Print sources do not move. Web sources do, and the numbers on this are not marginal.

Klein and colleagues (2014) examined references in arXiv, Elsevier and PubMed Central corpora and found link rot of 13%, 22% and 14% respectively for articles published as recently as 2012, rising sharply for older articles (s. 15). Coble and Karlin (2023) analysed 1,924 citations containing links in Digital Humanities Quarterly and found that 31% of them no longer worked correctly. In a Polish study of four library and information science journals, Roszkowski and Włodarczyk (2016) reported that over 70% of web citations were still accessible at the time of the study and estimated the half-life of a web citation at roughly seven years (s. 43).

Ott (2022) draws the distinction that matters for how you cite. Reference rot has two components: link rot, where the address no longer resolves, and content drift, where the address still works but the content behind it has changed, sometimes beyond recognition (s. 3). The article is a study of medical journal references, and the same mechanism applies to any web citation. Link rot announces itself with a 404. Content drift does not announce itself at all, which is exactly why a retrieval date and an archived copy are worth the thirty seconds they cost.

Archive the page, then cite it

Paste the URL into the Wayback Machine's "Save Page Now", or use Perma.cc if your institution provides it. You get a timestamped snapshot. Cite the original URL as normal, keep the snapshot link in your notes, and add it to the reference only if your supervisor or the journal asks for it. If the page vanishes between submission and defence, you have the page as it stood.

Do not let a chatbot type the URL

Ask an AI assistant to format a web citation and it will produce something that looks correct in every slot. The failure mode is well documented. Szeider (2025) summarises testing across several models, with fabrication rates for certain publication categories exceeding 80%, and notes that these fabrications include plausible but nonexistent author names, venues and DOIs (s. 2). A fabricated DOI is worse than no DOI, because it looks like the one thing in the citation that can be trusted.

The safe division of labour: you supply the real address, the tool formats it. Paste the URL into cytado's citation generator and it pulls the metadata from the page and the source databases rather than guessing at it, then builds the entry in the style you picked. For the whole list at once, the bibliography generator does the same job across every source in a document. If you want to check a reference someone else handed you, how to check if a source exists covers that separately.

Incidentally, the academic sources cited above were found through cytado's own source corpus, with the page numbers read off the printed pages. That is the same standard we would want from any citation, including ours.

The mistakes that cost marks

Before: www.who.int — the homepage, cited for a specific statistic. After: the full address of the page carrying that statistic, plus the page's own title and date.

Before: an access date written as "accessed 2024". After: a full date, because a year is not a snapshot.

Before: the URL as the whole citation, with no author, title or date. After: all five slots filled, with n.d. or the organisation name standing in where needed.

Before: a page number invented for a web page. After: no page number for an unpaginated page. Cite a paragraph number or a section heading if your style allows it. For PDFs and ebooks, where a page number does exist somewhere, where to get the page number explains which number to take.

Before: the same URL cited three times with three different access dates because it was checked on three different days. After: one date, the day you settled on the version you are quoting.

Fill the five slots, prefer an identifier over an address, note when you read it, and keep a snapshot. That is the whole method, and it is the same method in APA 7, MLA and Chicago, with only the punctuation changing between them.