A draft arrives with a polished reference list. Every journal article has a DOI—the familiar string beginning with “10.”—and every link resolves. The sources look official enough that the content moves straight to copyedit. But one reference is an opinion piece supporting a factual claim, another study tested a different population, and a third comes from a publication whose review practices are difficult to find. The links work. The evidence does not.
Crossref put this distinction in plain language during its August–October 2026 consultation on new member practices: a DOI confirms that an item was registered and provides metadata, but it is not a mark of quality or endorsement. The proposed practices are still under consultation, with publication planned for December 2026. The underlying lesson is already useful for anyone editing an article, course, clinical document, report, or policy: identifiers help you locate and trace evidence; they do not evaluate it for you.
The short answer
Treat a DOI as an address, not a trust badge. Check whether the source directly supports the claim, whether its methods and limits are visible, whether the publication explains its editorial practices, whether you have the right version, and whether a later reviewer can reconstruct your decision.
What a DOI proves—and what it does not
A digital object identifier gives a work a persistent identity. Following the DOI should lead to a landing page, and the associated metadata can identify the title, creators, publisher, date, funding, licence, references, and related updates when those fields have been supplied. Crossref’s public metadata service lets anyone retrieve the record for a registered DOI. That is valuable evidence that you have found the intended object rather than a stray PDF or a copied reference.
It does not prove that the work was peer reviewed, that the methods answer your question, that the conclusions are justified, that conflicts were managed, or that the publisher behaves responsibly. A DOI can identify a journal article, report, preprint, dataset, book chapter, correction, or other output. Those objects serve different purposes. Even complete metadata describes the work; it does not certify the truth of every sentence in it.
That is why “it has a DOI” is too small an answer to “can we rely on this?” The editorial question is not whether the source looks scholarly. It is whether this source is good enough for this claim, in this context, at this level of consequence.
A five-pass credibility check
Ask what the source actually needs to prove
Write the claim in one sentence, then name its type. Is it about what happened, why it happened, how common it is, what a rule requires, or what someone should do? A news release may establish that an agency announced a policy; it is weaker support for the policy’s exact legal effect. A review can summarize a field; an original study may be necessary for a precise result. Prefer direct, primary evidence when it is available.
Read past the title and abstract
Check who or what was studied, the comparison, the outcome, the timeframe, and the important exclusions. Look for uncertainty and limitations. The U.S. National Library of Medicine’s current journal-review criteria ask whether aims are clear, methods can be understood or reproduced, the design suits the aim, and the data support the conclusions. Those questions work at article level too.
Look for visible editorial practices
Can you find the owner, editorial board, peer-review process, ethics policy, conflicts, funding, fees, copyright terms, and correction process? The joint transparency principles from COPE, DOAJ, OASPA, and WAME make these features concrete. Their presence does not make every article sound; their absence removes useful ways to test the publication’s claims about itself.
Make sure the record matches the item
Resolve the DOI and compare the title, authors, date, and publication with the file in front of you. Note whether it is a preprint, accepted manuscript, version of record, correction, or update. Where available, Crossmark can expose corrections, retractions, and other changes. For a fuller status workflow, use our prepublication citation check.
Record why this source is suitable
Save the complete citation, link, access date when appropriate, version, pages or sections used, and one line on what the source supports. Record a material limitation: “adult hospital staff in one country,” “guidance, not binding law,” or “preprint, not peer reviewed.” This small note helps a reviewer check the claim without repeating the whole search and helps the next update find weak assumptions quickly.
Warning signs are prompts, not automatic verdicts
Be cautious when a DOI resolves to a page whose title or authors do not match the reference; a journal promises guaranteed acceptance or implausibly fast review; fees, ownership, editors, or review methods are hard to find; the article’s conclusion is much stronger than its results; or a precise claim relies only on an editorial, abstract, or secondary summary. The current ICMJE reference guidance recommends direct references to original research where possible and warns against citing predatory or pseudo-journals.
Do not replace judgment with a blacklist, an impact factor, or a database badge. ICMJE notes that there is no single validated mechanism for identifying deceptive journals. Conversely, a small, new, regional, or non-English publication is not unreliable merely because it is unfamiliar. Use transparent practices and the quality of the specific work as evidence. Escalate uncertainty when the claim could affect health, safety, compliance, money, or someone’s rights.
A practical example: one confident training claim
Suppose a compliance course says, “Microlearning improves knowledge retention by 50%,” and cites a paper with a DOI. The DOI resolves correctly. That completes the identity check, not the credibility check. The editor opens the paper and discovers a short, uncontrolled study of university students, a different definition of microlearning, and an immediate quiz rather than delayed retention. The article may be legitimate, yet it does not support the sentence as written.
The responsible fix might be to narrow the statement, replace the source with stronger evidence, or remove the percentage. The decision depends on the evidence, not on how polished the citation looks. A useful source can be unsuitable for a particular claim without being “bad research.” That distinction keeps editorial review fair and precise.
A 15-minute review for an existing source
- Minutes 1–3: isolate the claim. Highlight the exact words the reference is expected to support and classify the claim.
- Minutes 4–7: open the source. Confirm the item and version, then read the relevant results, guidance, or governing text in context.
- Minutes 8–10: inspect the limits. Compare population, setting, dates, definitions, methods, and uncertainty with the draft.
- Minutes 11–13: inspect the publication. Check review, ownership, conflicts, funding, ethics, and correction information in proportion to risk.
- Minutes 14–15: leave a trail. Record what the source supports, what it does not, and whether the wording, citation, or review level must change.
Not every citation needs fifteen minutes. A familiar agency page supporting a low-stakes definition may need a quick identity and currency check. A surprising medical statistic in patient material deserves deeper review by an appropriate subject-matter expert. Set the depth by consequence, novelty, and uncertainty—not by whether the reference arrived with a DOI.
The goal is not to make every editor a research-methods specialist. It is to stop the reference list from acting like decoration. A strong source trail shows where a claim came from, why the evidence fits, what its limits are, and what should trigger another look. The DOI helps keep that trail connected. Trust comes from the checks around it.
For editorial and content teams
Keep references connected to the claims they support.
Superscriptify helps teams clean, align, and reorder citations in Word, PowerPoint, eLearning exports, and structured publishing files—so human review can focus on whether the evidence is the right evidence.
Explore tools for publishersSources and further reading
- Crossref: Share Your Feedback on the New Crossref Member Practices, August 10, 2026
- Crossref: REST API metadata documentation
- Crossref: Crossmark content-status documentation
- U.S. National Library of Medicine: Journal Selection for MEDLINE, reviewed July 24, 2026
- COPE, DOAJ, OASPA, and WAME: Principles of Transparency and Best Practice in Scholarly Publishing
- International Committee of Medical Journal Editors: Preparing a Manuscript for Submission
- ICMJE: “Fake,” “Predatory,” and “Pseudo” Journals
This article provides general editorial and educational information, not medical, legal, research-integrity, or compliance advice. Source suitability depends on the claim, audience, jurisdiction, and consequences. Use qualified subject-matter, legal, statistical, or clinical review where the risk warrants it.