How to Find a DOI for an Article, Book, Dataset, or Preprint

How to Find a DOI: Quick Answer

Start on the work itself. Search the first PDF page, the publisher HTML, and the database Cite / Details pane for a string that begins with 10. and contains a slash. If nothing is printed, paste the title or a full reference into Crossref Metadata Search or Crossref Simple Text Query. For datasets, software, and arXiv preprints, search DataCite Commons as well — Crossref does not hold every DOI. Then resolve what you found at https://doi.org and confirm the landing page is the same work.

1. Search the PDF or HTML page for doi (header, footer, abstract, copyright line).
2. Open the full database record. PubMed, JSTOR, ScienceDirect, Wiley, SpringerLink, EBSCO, and ProQuest expose a DOI field when the publisher deposited one.
3. If nothing is printed, search Crossref for articles, books, and chapters. Search DataCite Commons for datasets, software, and many preprints.
4. Copy the string that starts with 10. and test https://doi.org/ plus that string. Strip any campus ezproxy host first.
5. If the resolver fails or lands on the wrong item, you do not yet have the DOI. Do not invent one.

What a DOI Is

The DOI Foundation governs the identifier and is the registration authority for ISO 26324, the Digital Object Identifier System (approved 2010, published 2012, revised 2022). Their definition is broader than "a journal-article URL":

That official example is the DOI Handbook itself. Prefix 10.1000 identifies the International DOI Foundation. Suffix 182 identifies the current handbook edition. The Foundation's handbook page (updated 10 December 2025) states that https://doi.org/10.1000/182 always identifies the latest version of record. The PDF is the version of record; the HTML is a reading copy. Prefacing any name with https://doi.org/ makes it actionable: the proxy looks the string up and redirects you to current information about the object.

The same handbook page notes that the text is under active review for ISO/DIS 26324, the draft expected to replace ISO 26324:2022. The identifier syntax you copy — 10. plus a slash — does not change while that review runs. What can change is handbook wording about registration-agency duties. Keep the live handbook URL, not a downloaded 2010s PDF, as your authority.

Resolution uses Handle System technology. Persistence is an organizational promise, not a file format: a publisher or repository that registers a DOI undertakes to keep the metadata pointed at the current location. That is why a 1953 Nature paper can still have a working DOI even though the journal's website has been rebuilt many times, and why a DOI is not the same thing as the URL in your browser's address bar.

You cannot mint a DOI yourself. Only DOI Registration Agencies — Crossref for most journal articles, books, and conference papers; DataCite for most datasets, software, and repository deposits; plus regional agencies such as JaLC, KISTI, Airiti, and mEDRA — assign names through their members. Pasting a made-up 10. string into a paper does not create an identifier. Crossref's public /works/{doi}/agency route will tell you which agency holds a name you already have: Niepel 2025 (10.1037/edu0000946) returns Crossref; Vaswani 2017 (10.48550/arXiv.1706.03762) returns DataCite; the handbook (10.1000/182) returns Public.

How to Recognize a DOI

Almost every scholarly DOI you will cite still starts with 10. and contains a slash. The part before the slash is the registrant prefix. The part after it is the suffix the registrant chose. Suffixes are arbitrary: they may look tidy (10.1037/edu0000946) or ugly (10.1175/1520-0442(1988)001<0715:COAPSR>2.0.CO;2). Both are real. Do not "clean up" punctuation inside a suffix.

The APA Style Common Reference Examples Guide (last updated 3 March 2026) uses that first row for Niepel, Hausen, Weber, and Möller's 2025 article in Journal of Educational Psychology. The DOI Foundation's own handbook example remains 10.1000/182. Do not treat 10.1000 as a journal prefix you can reuse for made-up articles.

Campus proxy wrappers are not part of the DOI. A link that looks like https://doi-org.ezproxy.college.edu/10.1037/edu0000946 is the same identifier with your library's login host spliced in. Before you cite it, rewrite the host to https://doi.org/10.1037/edu0000946. MLA Style Center makes the same point when a Project MUSE or Springer link arrives with a campus hostname: keep https://doi.org/ plus the 10. name.

DOI names are case-insensitive when resolved, but suffixes are stored as registered. Copy and paste; do not retype. The Crossref record for the Clow et al. 1988 Journal of Climate article that Crossref uses as its Simple Text Query demo stores the suffix in lowercase (…coapsr…) even though older print may have shown mixed case. Either string resolves. Guessing characters does not.

Why Finding the DOI Matters

A publisher URL can 404 the next time the journal migrates platforms. The DOI is the registered pointer that is supposed to survive that move. That is also why current citation manuals tell you to prefer a DOI over a database URL when both exist. Finding it is the research step; formatting it is a later step.

Chicago 18th edition (2024) makes the same display choice. CMOS 13.7 treats a DOI written as doi:10... as deprecated and wants the URL form. The live Notes-Bibliography Quick Guide shows Hyeyoung Kwon, "Inclusion Work," American Journal of Sociology 127, no. 6 (2022), as https://doi.org/10.1086/720277. CMOS 13.11 also tells you not to replace a DOI with a third-party shortener, including ShortDOI. Crossref's display guidelines (DOI 10.13003/5jchdy) reject a bare doi: prefix and the old dx.doi.org host, and they add this warning about ShortDOI:

APA's DOIs and URLs page (reprinted on the DePaul and UA Little Rock guides) adds two lookup rules students miss: include the DOI even if you read the print volume, and standardize older doi: or dx.doi.org printings to https://doi.org/ in the reference list. This page stops at finding and confirming the name. Period-after-DOI rules live on the usage guide.

Look on the Source First

Most of the time you do not need a search engine. The registrant already printed the DOI on the object. Open the PDF or the publisher HTML and use find-in-page for doi before you leave the file.

Journal article PDF and publisher page

On a Crossref-registered article the DOI is typically in one of four places. Check all four before you decide it is missing — some publishers put it only in the HTML "cite this" panel, and some PDFs print it only on page 1.

• First-page header or footer. Nature prints https://doi.org/10.1038/171737a0 on Watson and Crick's 1953 letter, even though the paper long predates the DOI system. Retrospective registration is common.
• Abstract / "About this article" block. APA Publishing, Wiley, Springer Nature, Elsevier, and the University of Chicago Press all put the DOI next to the title or CRediT line on the HTML version.
• Copyright or citation line. Often doi:10... in older PDFs, or a full https://doi.org/ link in files published after about 2017.
• "Cite this" or "Download citation" export. RIS, BibTeX, and the publisher's APA/MLA preview usually include a DO or doi field. Prefer that field over the database session URL sitting in the same export.

Books and chapters

Scholarly monographs and chapters that went through Crossref often carry a DOI on the title page verso, the chapter opener, or the publisher catalog record. An ISBN is not a DOI. Some publishers also register an ISBN-A (an actionable ISBN expressed as a DOI), but you should look the book up in Crossref rather than trying to construct that form. The APA 7 official book sample in the same 3 March 2026 guide is Levenson, Jinich, Vaz, and Rousmaniere, Deliberate Practice in Emotionally Focused Couple Therapy (2025), DOI 10.1037/0000436-000. If Crossref returns nothing, cite the book by ISBN and publisher data — see how to cite a book in APA or Chicago — and do not invent a 10. string from the ISBN digits.

Datasets, software, and preprints

Repositories that use DataCite — Zenodo, Figshare, Dryad, ICPSR, many university data archives — print the DOI on the landing page, usually under "Cite as" or next to the version number. Software releases deposited to Zenodo get a version DOI plus a concept DOI that always points at the latest version; copy the version you actually used. The APA 7 official dataset sample is William O'Donohue, Content Analysis of Undergraduate Psychology Textbooks (ICPSR 21600, Version V1), DOI 10.3886/ICPSR36966.v1. Preprint servers are mixed: bioRxiv and medRxiv register Crossref DOIs; arXiv assigns DataCite DOIs of the form 10.48550/arXiv. plus the arXiv id.

arXiv still accepts a separate journal-reference DOI when the paper is later published. That journal DOI and the 10.48550/arXiv. name are two identifiers for two objects. Cite the version you used. Replacing an arXiv version does not mint a new DataCite DOI; the same name keeps pointing at the latest /abs page.

Find a DOI in a Library Database

If you reached the article through a library database, the DOI is a metadata field, not the long URL that authenticated you. Open the full record — not just the search snippet — and look for a labeled DOI. Then resolve it on doi.org so you are not citing a proxy session.

APA 7 is explicit about the next decision. If the database record has a DOI, that is what goes in the reference — not the database name and not the permalink. If a widely available journal article has no DOI, APA's database-information page (same wording on the DePaul reprint) says to treat it like the print article and stop. MLA still wants a location: DOI, then permalink, then URL (How do I cite an article in a library database with no DOI?). Chicago prefers the DOI when one exists and otherwise a URL or the database name — the Quick Guide Hebert 1925 sample uses EBSCOhost because that record has no DOI. Those are formatting rules. The finding rule is the same in every style: look for the 10. field before you settle for a URL.

Look the DOI Up When It Is Not on the Page

Older scans, photocopies, and sloppy exports often drop the DOI even when one was later registered. Publishers have been minting retrospective DOIs for decades — Watson and Crick again — so "the PDF does not show one" is not evidence that none exists. Use a registration-agency search, not a generic web search, and match on title plus first author plus year before you accept a hit.

Crossref Metadata Search

search.crossref.org is the public catalog of metadata Crossref members have deposited: journal articles, books, chapters, grants, and more. Search the article title in quotation marks, or the first author plus a distinctive title word. Open the result and copy the DOI. This is a metadata search, not a full-text archive. A hit means a member registered that work with Crossref. It does not mean the PDF is free.

Crossref Simple Text Query

When you already have a formatted reference — from a bibliography, a syllabus, or a "cite this" box that omitted the DOI — paste the citation into Simple Text Query. Crossref's own documentation (last updated 6 October 2020; still the live page) describes the tool this way:

Crossref's worked example is a real 1988 paper. Paste it on one line, the way their docs show:

Clow GD, McKay CP, Simmons Jr. GM, and Wharton RA, Jr. 1988. Climatological observations and predicted sublimation rates at Lake Hoare, Antarctica. Journal of Climate 1:715-728.

That suffix is supposed to look hostile. It is the registered name. Simple Text Query accepts up to 1,000 references per submission, one per line or block, and can optionally return PubMed IDs beside the DOIs. Turn on "List all possible DOIs per reference" when a short title could match more than one paper, then compare authors and pagination yourself. The tool only searches Crossref deposits. It will not invent a DOI, and it will not see a DataCite-only dataset or an arXiv DataCite name.

Crossref REST API (no account)

The same Crossref corpus is queryable without a form. Crossref's REST API states that no sign-up is required. Two routes matter for lookup:

• `/works?query.bibliographic=` — paste a title, author, and year (or a full citation). The response is JSON metadata, not a formatted reference. Crossref's own documentation example is query.bibliographic=room+at+the+bottom with query.author=richard+feynman.
• `/works/{doi}/agency` — returns the registration agency for a name you already have. Official wording: "The registration agency for the specified DOI (e.g., Crossref or DataCite)." Use this when Simple Text Query is empty and you need to know whether to switch catalogs.

A bibliographic query that returns a high score is still a candidate, not a proof. Open the title and year in the JSON, then resolve the DOI. Content negotiation (Accept: application/citeproc+json on https://doi.org/10.…) is how citation managers and this site's DOI converters fetch metadata after lookup succeeds.

DataCite Commons

DataCite Commons is the public search for DataCite-registered works — datasets, software, samples, instruments, and many repository preprints — plus a subset of Crossref records that are linked to those works. DataCite Search was discontinued on 1 June 2023; do not use leftover search.datacite.org bookmarks. Commons is a discovery layer, not a place that mints identifiers; repositories register names in DataCite Fabrica.

Start in the Works tab. Search the title, the creator, or a known DOI. Crossref DOIs appear in Commons only when a Scholix literature–dataset link or a DataCite relatedIdentifier points at them, or because of older import patterns. A journal article that lives only in Crossref can be invisible here. The reverse is also true: Simple Text Query will not return 10.48550/arXiv.1706.03762.

Commons search uses OpenSearch query-string syntax. Hyphens, slashes, colons, and angle brackets inside a DOI are reserved characters. DataCite's own support page uses 10.7936/5bc9-0k33 as the plain example and then shows why a quoted search is safer:

"10.21938/W:hWbPah3wBabuqewQULTA"

doi:10.2314/KXP\:1738669661

If a dataset DOI you copied from a landing page "is not in Commons," quote the whole 10. string before you conclude it was never registered. A Draft or Registered (not Findable) DataCite name will also stay hidden.

PubMed first, then the PMC ID Converter

Biomedical papers often travel with three identifiers: a DOI, a PMID, and sometimes a PMCID. They are not interchangeable. In PubMed, the DOI appears on the abstract page when NCBI has one. Watson and Crick's 1953 letter is PubMed 13054692; Europe PMC lists DOI 10.1038/171737a0 on the same record and reports the article is not in PMC. You can also search a known DOI directly.

The converter itself moved. Use the PMC ID Converter and its API documentation (last modified 10 June 2025). Older ncbi.nlm.nih.gov/pmc/pmctopmid/ bookmarks redirect there. NCBI is explicit about the scope:

A PMID in a Vancouver or NLM reference is not a substitute for the DOI in APA, MLA, or Chicago. Use the PubMed or Europe PMC abstract when the paper is not in PMC. Watson and Crick is the worked case: PMID 13054692 has a DOI, and the converter will still return nothing because inPMC is N.

Which registry? Crossref miss is not "no DOI"

A string that starts with 10. is registered with one agency, not with every agency. Crossref Simple Text Query will not return 10.48550/arXiv.1706.03762 because that name lives in DataCite. The reverse is also true: most journal articles never appear as primary records in DataCite Commons. If one catalog is empty, search the other before you stop. When you already have a 10. string, /works/{doi}/agency is the faster test.

• Crossref first for journal articles, books, chapters, and most conference papers.
• DataCite Commons first for datasets, software, samples, and arXiv e-prints.
• PubMed or Europe PMC when you already have a PMID. Use the PMC converter only if the article is in PMC.
• doi.org last to prove identity. A 404 there means the string is not a registered DOI, not that the PDF is paywalled.

Google Scholar — useful, not authoritative

Google Scholar is a discovery layer, not a registration agency. The quotation-mark Cite box sometimes includes a DOI and often does not. The snippet under the title almost never does. Click through to the publisher or repository page, or take the title into Crossref. Scholar also conflates versions: a preprint, a journal article, and a chapter can share a result cluster. Match the year, the container title, and the page or article number before you copy a 10. string from a third-party HTML copy. For library-subscribed full text, start in a library database whose record already has a DOI field.

A Lookup Path by Source Type

How to Test That You Found the Right DOI

A string that starts with 10. is not automatically a DOI for this work. Resolve it.

1. Go to https://doi.org and paste the name, with or without the https://doi.org/ prefix.
2. Confirm the landing page is the same work: same authors, title, year, and container. A 403 or publisher login wall can still be a correct DOI — access is not the test, identity is. Niepel 10.1037/edu0000946 and Kwon 10.1086/720277 both resolve and then often 403 from this network; that is still a successful identity check.
3. If you land on a different article, a journal homepage, or a generic 404, you copied a neighbor's identifier or a broken suffix. Go back to Crossref or DataCite and match the title.
4. If you need to know which registry holds the name, request https://api.crossref.org/works/ plus the DOI plus /agency.
5. If you need machine-readable metadata, request it with content negotiation (Accept: application/citeproc+json on https://doi.org/10.…). Citation managers and this site's DOI converters do that for you.

Some DOIs use multiple resolution and can offer more than one URL (publisher page plus a repository copy). The Foundation's resolver picks the appropriate target. You still cite the DOI, not whichever URL it happened to send your browser to today.

DOI vs PMID, PMCID, ISBN, and arXiv ID

Students often paste the first alphanumeric string on the page into a citation generator and call it a DOI. These identifiers live next to each other and do different jobs.

The glossary entries for DOI, PMID, ISBN, and arXiv id keep the short definitions. The practical test remains: if it does not start with 10. and resolve at doi.org to this work, it is not this work's DOI.

What If There Is No DOI?

Plenty of citable sources never received one: most newspapers and magazines, ordinary websites, many books from trade publishers, student papers, and articles that predate retrospective registration and were never deposited. Grey literature is mixed — see grey literature. Absence after a real search is normal. Absence because you only glanced at page 2 of the PDF is not.

• Search Crossref (and DataCite, if the object is a dataset or an arXiv preprint) before you stop. Retrospective DOIs are easy to miss on a scanned PDF.
• APA 7, widely available journal from a database, no DOI: end the reference after the page range. Do not add the database URL. The DePaul reprint of APA's database page states this next to the include-the-DOI rule.
• APA 7, online work with a URL that will work for readers and no DOI: use the URL. Do not pair it with a fake DOI.
• MLA 9: DOI, then permalink, then URL (Handbook 5.84). Database articles without a DOI may still take a permalink that leads to gated content.
• Chicago 18: prefer the DOI; otherwise a URL or, for some library sources, the database name (Quick Guide Hebert / EBSCOhost). Do not use a bit.ly or ShortDOI stand-in.
• Books without a DOI: use the ISBN and publisher. An ISBN is enough for ISBN-to-APA conversion; it is not a DOI.
• Websites: they almost never have DOIs. Cite the page. See How to Cite a Website in APA or MLA.

Common Lookup Mistakes

• Inventing a DOI from an ISSN, ISBN, or page number. Registration agencies assign names. Pattern-matching 10. plus a journal code is how fabricated references get into papers.
• Copying the campus ezproxy host. Rewrite to https://doi.org/10.… so a reader outside your campus can resolve it.
• Treating a PMID or an arXiv id as a DOI. Convert, then cite the DOI.
• Stopping at Crossref when the object is a dataset or an arXiv preprint. Search DataCite Commons.
• Searching DataCite Commons without quoting a hyphenated or slashed DOI. OpenSearch treats -, /, and : as operators. Quote the whole name.
• Using the PMC ID Converter on a paper that is not in PMC. Watson and Crick 1953 will return nothing even though Europe PMC already lists the DOI.
• Stopping at Google Scholar's Cite box. It is incomplete more often than not. Finish in Crossref or on the publisher page.
• Using `10.1000/…` as a journal-article example in your own paper. 10.1000/182 is the DOI Handbook. It is not a template for an article you read.
• Citing ShortDOI (`10/abcd`) or a bit.ly link. Crossref says ShortDOIs are not really DOIs. CMOS 13.11 rejects third-party shorteners. APA and MLA want the standard https://doi.org/10.… form.
• Assuming a 403 means the DOI is wrong. Identity and access are different questions. Check the landing-page bibliographic data.
• Assuming a PDF without a printed DOI has none. Run Simple Text Query on the reference you already have.
• Using a neighbor article's DOI. Conference special issues and "related articles" sidebars are designed to be clicked by mistake. Resolve and read the title.

After You Have the DOI

Once the identifier resolves to the right work, you are done with lookup. Paste the DOI into the citation generator, convert a list with DOI to APA, DOI to MLA, or DOI to BibTeX, or import a RIS file into Zotero or EndNote. Worked reference punctuation — hanging indents, https://doi.org/ vs a bare 10., the extra period MLA still places after the URL — belongs in How to Use DOIs in Research and in the journal-article guides for APA, MLA, and Chicago.

Official documentation worth keeping open while you look things up: the DOI Foundation explainer, the DOI Handbook (10 December 2025), Crossref Simple Text Query, the Crossref REST API, Crossref display guidelines, DataCite Commons search help, arXiv assigned DOIs, the PMC ID Converter API, APA Style Common Reference Examples (3 March 2026), and the MLA Style Center note on database articles with no DOI.

Related guides and tools

• How to Use DOIs in Research
• How to Cite a Journal Article in APA
• How to Cite a Journal Article in MLA
• How to Cite a Journal Article in Chicago
• How to Cite PDF Documents
• How to Cite a Dataset in APA
• How to Use Google Scholar
• Library Databases for Research
• DOI to APA converter
• DOI to BibTeX converter
• How to cite a journal article in APA (examples)
• DOI in the glossary