Let’s cite!
Paste a DOI, ISBN, PubMed or arXiv ID, a web address, or just a title. Your list is saved in this browser, so you can close the tab and come back to it.
Formatted by citeproc-js © Frank Bennett, unmodified, under the CPAL — citeproc-js implements the Citation Style Language. Styles from the CSL project, CC BY-SA 3.0 (our copies).
A citation generator turns a source into a correctly formatted reference. You paste what you already have — a DOI, an ISBN, a PubMed or arXiv ID, a web address, or just a title — and it finds the publisher’s own record, formats it in the style your department asks for, and gives you the in-text citation to go with it. The whole list downloads as a Word file with the hanging indent, spacing and alphabetical order already set.
It supports APA 7, MLA 9, Chicago, Harvard, IEEE, Vancouver and 10,855 more, across thirty-six kinds of source. It is free for everyone, with no sign-up and nothing held back for a paid plan, because the formatting runs in your browser rather than on our servers.
The section worth reading: why almost every citation generator gets titles wrong, and what this one does about it.
Three steps, about twenty seconds a source.
Eighteen styles sit one click away in the picker at the top. If your department or journal uses something else, search the full Citation Style Language repository — 10,861 styles, including most journals and a great many individual universities.
A DOI, an ISBN, a PubMed ID, an arXiv ID, a link, or a title. The tool works out which one it is and asks the right catalogue — Crossref, OpenLibrary, PubMed, DataCite or Semantic Scholar. Nothing found? The manual form covers 36 source types.
Copy it with its formatting and paste into Word or Google Docs, or download a .docx that already has the hanging indent, the spacing and the right heading. BibTeX and RIS are there too, for Zotero, Mendeley, EndNote and LaTeX.
There is no drop-down asking what you are about to paste. The tool reads it and decides, then labels the entry with the catalogue the details came from.
Most free tools hand you the reference list and leave the harder half — the bit you actually type into a sentence — as an exercise for the reader.
It is almost never the punctuation. It is the capital letters, and it is wrong on every journal article in the list.
It still cannot know that War is part of a name in World War II and not in the war years. No software can. What it can do is stop pretending the problem does not exist.
Eighteen styles are pinned at the top of the picker. The rest of the Citation Style Language repository is a search box away, fetched only when you choose one.
The difference between a citation string and a finished reference page.
Two papers by the same authors in the same year become 2020a and 2020b, in the list and in the in-text citation, exactly as APA requires. Numbered styles renumber themselves as you add and delete. This sounds obvious; several well-known tools get it wrong.
The .docx download carries the hanging indent, the line spacing and the alphabetical order the style calls for, under the right heading. Copying keeps the italics when you paste into Word or Google Docs. Most free tools hand you a blob of text and leave the formatting to you.
Each entry is labelled Crossref, OpenLibrary, PubMed, DataCite, the page itself, or you. Metadata scraped from a page is marked as worth checking, because it usually is, and missing required fields are counted on the card rather than left for a marker to find.
Sentence case, DOI form, page ranges, word editions such as “Fourth edition.” reduced to a plain number — each fix is listed on the source it changed, with the original text and a one-click undo. Nothing is altered silently.
Your list stays readable, editable and exportable, free, for as long as you keep it, and it lives in your browser rather than on our servers. We are not going to put a citation you already made behind a paywall, which is the loudest complaint about several of the big names in this category.
There is no AI in this tool. It never touches the daily words, the per-request cap or the requests-per-minute limit that the AI tools share, so it behaves identically for a guest, a free account and a Pro subscriber.
Each one mapped to the correct underlying type, so the style renders it the way it is meant to.
Journal article, thesis or dissertation, conference paper, preprint, unpublished manuscript.
Book, book chapter, edited book — with editors, translators and editions handled separately.
Website or web page, blog post, social media post, and generative AI tools such as ChatGPT.
Newspaper, magazine, online video, film, TV episode, podcast, song, artwork, map, review.
Software or code and datasets, with version numbers and DOIs where they exist.
Report, government document and standard, each with its issuing body and number.
Interview, speech or lecture, and personal communication — which APA cites in the text only.
Legal case, statute, bill and patent, plus encyclopedia and dictionary entries.
Your department’s handbook is the authority. This is the usual shape of it.
| Style | Usually used for | In text it looks like |
|---|---|---|
| APA 7 | Psychology, education, nursing, business, most social sciences | (Smith & Doe, 2020, p. 23) |
| MLA 9 | English, literature, languages, cultural studies | (Smith and Doe 23) |
| Chicago (notes) | History, art history, theology — footnote based | A numbered footnote |
| Chicago (author–date) | Sciences and social sciences using Chicago | (Smith and Doe 2020, 23) |
| Harvard | Widely used across UK universities, with local variations | (Smith and Doe, 2020, p. 23) |
| Vancouver / NLM | Medicine, nursing, biomedical sciences | [4] |
| IEEE | Engineering, electronics, computer science | [4] |
| AMA 11 | Medical journals following the AMA manual | A superscript number |
| OSCOLA | UK law | A numbered footnote |
Not sure? Pick the one your handbook names, generate the list, and switch styles later — the sources are stored as data, so changing style re-formats everything rather than making you start again.
A real lookup, unedited, including the correction the tool made and reported.
10.1126/science.1202775
Nothing else. No source type to choose, no fields to fill, no account. An ISBN, a PubMed ID, an arXiv ID, a link or a plain title goes in the same box and takes the same one step.
Golder, S. A., & Macy, M. W. (2011). Diurnal and seasonal mood vary with work, sleep, and daylength across diverse cultures. Science, 333(6051), 1878–1881. https://doi.org/10.1126/science.1202775
(Golder & Macy, 2011)
Title converted to sentence case. Was: “Diurnal and Seasonal Mood Vary with Work, Sleep, and Daylength Across Diverse Cultures”.
Four things this tool cannot do for you, said plainly.
Yes, completely, for everyone. There is no sign-up, no ads on this page, nothing held back for a paid plan, and no cap on how many sources you add. The formatting happens in your browser and costs us nothing to run, so there is no reason to charge for it — and no reason to lock a citation you have already made.
Eighteen are one click away: APA 7 and 6, MLA 9, Chicago 18 in both its author–date and notes-and-bibliography forms, Harvard (Cite Them Right 12), IEEE, Vancouver, AMA 11, ACS, CSE in two variants, ASA, APSA, ACM, OSCOLA, Nature and Elsevier Harvard. Behind those sits the whole Citation Style Language repository — 10,861 styles, including most journals and a great many individual universities — searchable by name.
A DOI, an ISBN, a PubMed ID, a PubMed Central ID, an arXiv ID, a web address, or just the title of an article or book. The tool works out which of those you gave it and asks the right catalogue. If nothing matches, the manual form is one click away and produces exactly the same result.
Crossref for DOIs and article titles, OpenLibrary for ISBNs and books, PubMed for medical articles, DataCite for datasets and preprints, Semantic Scholar for arXiv papers, and a page's own published metadata for a plain web address. Every entry is labelled with the catalogue it came from, so you can see how much to trust it.
Yes — every entry shows its in-text form directly underneath, with its own copy button. You can add a page number or one of nine other locator types, and tick a box if you have already named the author in your sentence, which switches the citation to its narrative form. In footnote styles such as Chicago notes-and-bibliography and OSCOLA, that line is the full first footnote rather than a parenthetical.
Because the style requires it. APA, AMA, Vancouver, Harvard and several others print article titles in sentence case, while Crossref and most publishers store them in title case. Nothing in the citation software converts between the two, so we do it — and then tell you, entry by entry, exactly what changed, with a button to put your wording back. Proper nouns are the one thing no machine can reliably spot, which is why the change is flagged rather than made quietly.
Yes. Your list is kept in your own browser's storage, so a refresh, a crash or a closed tab will not lose it, and it is still there tomorrow. It never reaches our servers. The flip side is that it lives on that one browser: clearing your browsing data or moving to another device will not carry it across, so export anything you need to keep.
Yes. The Word file comes out with the hanging indent, line spacing and alphabetical order the style calls for, headed References, Works Cited or Bibliography as appropriate, so it can go straight into your assignment. You can also copy the list with its formatting intact and paste it into Word or Google Docs, or export RTF, BibTeX, RIS, HTML, CSL-JSON or plain text.
Thirty-six. Journal articles, books, book chapters and edited books; websites, blog posts, social media posts and generative AI tools; newspaper and magazine articles, online video, film, TV, podcasts, music, artwork, maps and reviews; software and datasets; theses, conference papers, preprints and unpublished manuscripts; reports, government documents and standards; encyclopedia and dictionary entries; interviews, lectures and personal communications; and legal cases, statutes, bills and patents.
No. The formatting engine runs inside your browser and your list stays in your browser's storage. The one exception is citing a plain web address: a browser is not allowed to fetch other people's pages, so that single request goes through our server, which reads the page's published metadata and returns it. We do not store the page, and we do not keep your list.
The formatting comes from citeproc-js and the official Citation Style Language style files — the same engine and the same styles Zotero and Mendeley use — so the shape of the entry is as correct as the style file itself. What no tool can guarantee is the data going in: a publisher record with a missing issue number or a mangled author name produces a tidy citation of the wrong facts. We show which fields are missing and where each field came from, but the last read-through is yours.
No. There is no AI in this tool and no metering on it. It never touches the daily words, the per-request limit or the requests-per-minute limit that the AI tools share, on any plan. It behaves the same for a guest, a free account and a Pro subscriber.
Cite your sources here, then work on the writing around them. Drafts you make in the AI tools are kept in your drafts.
Upgrade
