What is a DOI and how to use one

Somewhere near the top of almost every paper you will read for your thesis sits a short string that starts with the number 10 and a dot. Most students copy it without thinking, some ignore it entirely, and a surprising number lose hours or marks because of it. That string is a DOI, and it is the single most useful piece of metadata a paper has.

A DOI, or Digital Object Identifier, is a persistent identifier for a piece of scholarly output. Unlike a URL, which breaks when a publisher redesigns its site or a journal changes hands, a DOI is designed to keep resolving to the current location of the work for as long as the record exists.

This guide explains what a DOI is and is not, how to read one, where to find it when it is hiding, and how to put it to work: in citations, in your reference library, and in your final bibliography.

What a DOI is (and what it is not)

The DOI system is coordinated by the International DOI Foundation and operated through registration agencies. Crossref handles most journal articles and scholarly books; DataCite covers datasets, preprints and other research outputs. When a publisher registers a DOI, it takes on the responsibility of keeping the identifier pointed at the work, wherever the work later moves.

It is just as important to be clear about what a DOI is not:

• Not a mark of quality. Weak and predatory journals register DOIs too. A DOI tells you a record exists, not that the work is sound.
• Not proof of peer review. Preprints, reports, theses and datasets all carry DOIs without having passed review.
• Not a URL. A DOI can be expressed as a link, but the identifier is the string itself, and it outlives any particular web address.
• Not universal. Older papers, some conference proceedings and many book chapters never received one, and that is normal.

How to read a DOI: prefix and suffix

Every DOI has two parts separated by a slash. The prefix, which always starts with 10 followed by a dot, identifies the registrant, usually the publisher. The suffix is whatever string the publisher chose for that specific work: sometimes structured, sometimes opaque.

Take a made-up example: 10.1234/jhe.2026.0187. The 10.1234 part points to the registrant, and jhe.2026.0187 is the article's own tag. You never need to decode a suffix. The whole string is the identifier, it is not case-sensitive, and the convention is to write it in lower case.

To resolve a DOI, put it after doi.org in your browser's address bar. The resolver forwards you to the current landing page for that work. This is also why modern citation styles present DOIs as resolvable links rather than as plain strings: the reader can follow them directly to the source.

How to find the DOI of a paper

Publishers place DOIs in a few predictable spots, so a missing DOI is usually a looking problem rather than an absence:

• On the first page of the PDF, usually in the header or footer, often near the copyright line or the article history dates.
• On the article's landing page, typically under the title or inside a cite, share or export panel.
• In the database record: OpenAlex, PubMed, Semantic Scholar and arXiv listings all expose the DOI where one exists.
• In another paper's reference list, since many journals now print DOIs for every entry they cite.

What if a paper has no DOI?

If you have the paper in front of you and cannot find a DOI anywhere, search the exact title in a scholarly index and check the authoritative record. Sometimes the published version has a DOI while the copy you hold is an earlier preprint or an author manuscript without one.

If nothing surfaces, the work may genuinely not have a DOI. That does not make it uncitable. Cite it by its normal bibliographic details, exactly as your citation style prescribes for the source type, and move on. A DOI is a convenience and a safeguard, not a requirement for a source to exist.

Be especially careful with the preprint case. If you read the preprint but the published version is what your field considers citable, decide deliberately which version you are citing, and use that version's identifier and page numbers. Mixing metadata from two versions of the same work is one of the most common reference list errors examiners catch, because the year, title and pagination quietly stop agreeing with each other.

How to turn a DOI into a full citation

The fastest use of a DOI is generating a complete, correctly formatted reference. Because a DOI resolves to authoritative metadata (authors, title, journal, volume, issue, pages, year), a citation built from a DOI is far more reliable than one typed by hand or scraped from a PDF's text layer.

CiteDash offers free tools for exactly this. The DOI lookup tool takes a DOI and returns the paper's full details, and the free citation generators turn a source into a formatted reference in any of 12 styles, from APA and Harvard to IEEE, Vancouver and Chicago.

Citing from the DOI rather than retyping details removes a whole class of bibliography errors: transposed page numbers, misspelled author names, missing issue numbers, and the wrong year pulled from a preprint version of the paper.

Should the DOI appear in your reference list?

Most current citation styles ask for the DOI in the reference entry when one exists, and several present it as a resolvable link. The details differ by style and by edition, so check the version your department mandates rather than copying whatever format you saw in a published paper, which may follow an older edition.

The practical rule: capture the DOI for every source regardless of whether your style prints it. Styles change, departments differ, and a bibliography with DOIs attached can be reformatted for any requirement. One without them cannot be upgraded automatically.

There is also an examiner-facing benefit. A reference list where every entry carries a working DOI signals care, and it lets a sceptical reader verify any source in one click. Examiners spot-check references more often than candidates expect, and the smoothest possible outcome of a spot-check is that the link resolves instantly to exactly the paper you described.

Using DOIs to build a clean reference library

DOIs matter most at the library level. When every source in your reference library is keyed to a DOI, duplicates become obvious, metadata stays consistent, and everything downstream (your citation style, your exports, your submission checks) inherits clean data.

In the CiteDash Library, you can fetch a paper's full details from a DOI alone: paste the identifier and the record is created from authoritative metadata rather than from whatever a PDF's text layer happens to contain. Keying records to identifiers like this is also what makes reliable retraction badges possible later, because status checks need to know exactly which paper they are checking.

Identifier keying is the backbone of deduplication too. Two records with the same DOI are the same paper, no matter how differently two databases have abbreviated the title or formatted the author list. Anyone who has merged a supervisor's reading list into their own library knows how much time that one equivalence saves: instead of eyeballing near-identical titles, you compare identifiers and the duplicates announce themselves.

When a DOI does not resolve

Occasionally a DOI will fail to resolve. Before assuming the record is broken, work through the common causes:

• Trailing punctuation: a period or closing bracket copied from the end of a reference list entry will break resolution. Strip anything after the suffix.
• Line-break hyphens: a DOI split across two lines in a PDF sometimes gains a hyphen that is not part of the identifier.
• Very new papers: a DOI can appear on an online-first article shortly before registration propagates. Try again later.
• Genuinely wrong DOIs: reference lists contain typos like everything else. Search the title instead and take the DOI from the authoritative record.
• Paywall confusion: a DOI that resolves to a page asking for payment has still resolved correctly. The identifier locates the record; access is a separate question, and your library or an open-access version may get you the full text.

DOI vs PMID, arXiv ID and ISBN

A DOI is not the only identifier you will meet. PubMed assigns PMIDs to biomedical papers, arXiv assigns its own identifiers to preprints, and books carry ISBNs. These coexist with DOIs rather than compete with them: one paper can hold a PMID, an arXiv ID and a DOI at the same time, each keying it into a different system.

When you have a choice, prefer the DOI for citation purposes. It is the identifier citation styles expect, the one publishers maintain, and the one most tools use to fetch metadata. Keep the others as useful aliases for searching within their home databases.

Using DOIs to search and screen literature

A DOI is also the fastest way to run down a specific paper you already know about. When a supervisor emails you a reference, or you spot a promising source in another paper's bibliography, searching by DOI takes you straight to the exact record with none of the ambiguity of title searching, where similar titles, republished versions and conference-versus-journal variants muddy the results.

This is how the CiteDash Literature Finder handles known-item lookups too: paste a DOI and it resolves the exact paper across an internal corpus and live scholarly indexes including OpenAlex, PubMed, Semantic Scholar and arXiv. The identifier does the disambiguation for you, so screening a reading list of twenty inherited references becomes a paste-and-check exercise rather than twenty separate judgement calls about whether you found the right paper.

Make the DOI your primary key

Treat the DOI as the primary key of your research workflow: capture it at the moment of discovery, cite from it rather than from memory or from a PDF, and let it keep your library and bibliography consistent from first search to final submission.

The habit costs seconds per paper. It pays back every time a style changes, a duplicate needs merging, a source needs verifying, or an examiner follows a reference to check your reading. Every one of those moments goes better when the identifier is already there.