Check that a reference DOI leads to the work you actually used
Check that a reference DOI identifies the work you consulted, resolve ambiguous matches and separate access problems from formatting errors before submission.
About 7 min read · Allow extra time for the practical checks
Open each DOI and compare the destination with the source you consulted before treating the reference as checked. A link that opens successfully can still identify a different article, a chapter instead of a whole book, or a version you did not use. Check source identity first, then tidy the link’s presentation.
This is a focused final-reference audit. It complements matching in-text citations with the reference list: that check establishes correspondence inside your document; this one checks where the reference sends a reader.
Start with the source beside you
Keep the actual article, chapter or other work open alongside its reference entry. Compare its title, authors, publication and identifying details with the destination reached through the DOI. Record a discrepancy rather than correcting from memory. A similar title is a reason to investigate, not enough to establish a match.
For a chapter, inspect the chapter identity as well as the containing book. For an article, check whether you have the article itself or a separate notice about it. If the record describes another object, do not silently replace the reference details to make them fit the link. Return to the work that supported your writing.
These comparisons are a practical audit method, not a claim that every publisher landing page presents the same fields. Where information is incomplete or contradictory, retain the uncertainty in your working record until you can resolve it.
Use a DOI search as a candidate check
If the DOI is missing from your notes, check the publication’s own record. Crossref’s Simple Text Query guidance also describes finding DOI matches from bibliographic references. It explains that ambiguous or incomplete metadata can produce multiple candidates requiring evaluation.
Do not paste the first returned identifier into your paper simply because a search found it. Compare the candidate with the source already in hand. If several candidates remain plausible, leave the entry unresolved while checking the publisher record or asking a librarian for help.
A search that does not find a match does not give you permission to construct one. Do not borrow an identifier from a nearby reference or guess its suffix from the journal’s other articles.
Compare three plausible matches before selecting one
A successful search can return records that share some words without identifying the same work. Use the item you actually consulted to distinguish a chapter from its containing book or a separate article.
The following exercise is wholly fictional. All people, titles, dates, page ranges and candidate labels are invented. Candidate A, B and C are working labels, not DOI strings, and none represents a tested real link.
Suppose your notes contain only “Ng, 2024, Designing a Study.” You still have the chapter PDF. Its opening page identifies R. Ng and S. Osei as the chapter authors, Designing a Study as the chapter title, and Planning Applied Research as the containing book. The chapter runs from pages 71–88 and gives 2024 as its year. These details belong to this invented exercise; do not reuse them as a real reference.
You compare three candidate records:
| Candidate | What its record describes | Comparison with the consulted chapter |
|---|---|---|
| A | A chapter by R. Ng and S. Osei titled Designing a Study, in Planning Applied Research, pages 71–88, 2024 | The available identity details agree; keep it for a final destination check |
| B | The whole book Planning Applied Research, edited by M. Patel, 2024 | The containing title agrees, but the record describes the book rather than the chapter |
| C | A journal article titled Designing a Study by J. Hart, in Fieldwork Methods, 2024 | The short title and year agree, but the authors and publication do not |
Candidate B might look convincing if you search only the book title. Candidate C might look convincing if you rely on your abbreviated notes. Returning to the consulted chapter supplies the details needed to distinguish them.
Candidate A is the next one to verify, not a link you should now claim to have tested. In real work, open its actual DOI destination and confirm that it identifies the consulted chapter. If the destination omits a detail you need, check the publisher record or seek help resolving the discrepancy. Do not invent an identifier because the hypothetical table makes the answer seem obvious.
Keep a short decision note: “Chapter record selected for checking; whole-book and different-author records rejected.” Then record what the final check established. This preserves the reason for your choice if a reference manager later imports a different record.
The exercise also shows why you should not repair an incomplete reference by copying everything from the first search result. You could produce a perfectly formatted entry for the wrong object. Recover the consulted work’s identity first, then apply the required reference format.
Separate identity, access and formatting
Use separate notes for three questions: does the link identify the correct work; can you reach its record; and is the link presented correctly? Treating these as one tick box hides useful distinctions.
The following audit is wholly fictional. The reference labels and situations are invented; no real DOI, author or publication is implied.
| Working entry | What the check finds | Decision before submission |
|---|---|---|
| Reference A | Link opens, but the destination has a different title and author group | Do not mark it verified; locate the identifier belonging to the consulted source |
| Reference B | Destination identifies the consulted chapter; full text asks for institutional access | Record the identity match separately from access; do not substitute an unrelated freely available work |
| Reference C | DOI copied from the publication includes a manual line break in the draft | Compare with the original string, repair the transcription and test again |
| Reference D | Two search candidates remain plausible because the working reference lacks chapter details | Recover the chapter details before selecting either candidate |
| Reference E | Working notes show a downloaded report, but the reference points to an underlying dataset | Check what supported the sentence; do not imply analysis of data you never used |
The ledger makes the next action visible. For Reference A, changing the display style will not repair the identity mismatch. For Reference C, a transcription repair may be sufficient, but the destination still needs checking afterward. Reference E calls for a statistic-versus-dataset source decision, not a more impressive-looking identifier.
Apply the required reference style
APA’s DOI and URL guidance requires an available DOI for a work even when you used its print version. For an online work with both a DOI and a URL, APA uses the DOI alone. Its rules for online material without a DOI depend on the source type; do not apply a blanket instruction to add every database address.
APA presents a DOI as a link beginning with https://doi.org/, followed by the identifier. Copy the checked string, preserve its punctuation and capitalization, and avoid manual line breaks or a final full stop. Crossref’s display guidelines also recommend the full HTTPS resolver form for Crossref DOIs. An older display format does not by itself establish that an identifier is wrong.
Follow the required style for the complete reference, rather than replacing its author, date or title with a link. Formatting cannot establish that the source supports your sentence or that you actually read it.
Check the submitted file
After your last reference update, test the links in the file you intend to submit. Inspect both the visible text and the destination: a word processor can display one string while retaining another hyperlink target. If exporting a PDF, include that export in the check rather than assuming the editable document proves it works.
Keep a short unresolved list containing the entry, observed problem and next action. Clear it through evidence, not by deleting awkward links or marking them checked. If an identifier remains uncertain, seek help locating the source record and follow the applicable reference guidance. Never manufacture a successful audit.
