Skip to main content

Literature Guide

How to Verify a Citation

How to verify a citation properly: resolve the identifier, confirm the record, check for retractions, and establish whether the source supports the claim.

[Research Use Only] This guide is for controlled laboratory research workflows only. It is not for human or veterinary use and does not provide applied-use guidance.

Two questions wearing one coat

A citation makes two implicit promises. The first is that the referenced work exists as described. The second, much larger, is that the work supports the statement it has been attached to. Almost all citation checking addresses the first and stops. That is understandable — the first is quick, mechanical and satisfying — but a citation can be entirely real, correctly formatted, perfectly resolvable, and still not say what the sentence citing it claims. The documentation is the evidence, and reading it is the only way to know what it establishes.

The verification chain

Working through it in order matters, because each step narrows ambiguity that the next step would otherwise inherit. Skipping to the source document without an identifier is how people end up reading a different paper with a similar title.

  1. ClaimFix exactly what is being asserted — including numbers, populations, conditions and any hedging.
  2. CitationLocate the reference the claim points to. A claim with no citation cannot be verified, only believed.
  3. IdentifierExtract a DOI, PMID or PMCID. Titles are ambiguous; identifiers are not.
  4. Authoritative recordResolve the identifier at doi.org or PubMed — not at the page making the claim.
  5. Source documentOpen the article itself, rather than its abstract or someone’s summary of it.
  6. Relevant passageFind the specific result, table or statement the claim actually rests on.
  7. JudgementDecide whether that passage supports the claim as stated, supports something narrower, or does not support it.
The first five steps establish that the citation exists. Only the last two establish that it is doing the work the claim requires of it.

What each identifier is good for

The three identifiers a citation may carry are not interchangeable, and knowing which one you have determines where it resolves and what it proves.

IdentifierWhat it identifiesWhere it resolvesWhat it does not tell you
DOIA specific object, persistently, under ISO 26324doi.orgWhether the work is sound, current, or free of corrections
PMIDA record in PubMedpubmed.ncbi.nlm.nih.govWhether full text is available anywhere
PMCIDA full-text copy held in PubMed Centralpmc.ncbi.nlm.nih.govThat this is necessarily the publisher’s version of record
A citation carrying none of these is not automatically wrong, but it is materially harder to check — and being hard to check is itself worth noting.

Why a DOI is the strongest starting point

The DOI Foundation describes a DOI as a persistent identifier — a long-lasting reference to an object or to information about an object — conforming to ISO 26324. Resolution looks up the identifier and redirects to the object it points to, and because metadata and services associated with a DOI can be updated without changing the DOI itself, the identifier survives the object moving. The foundation is candid about where that durability comes from: persistence is a function of organizations, not of technology, requiring a persistent organisation, agreed policies and defined processes.

  • Prefix and suffix separated by a forward slash, resolving via doi.org
  • Designed to outlive changes in the location of the content
  • Resolves to a current location, which may differ from the one originally published
  • Guarantees identity and persistence — not quality, currency or correctness

Checking the record field by field

Once resolved, compare the record against the citation as given. Mismatches are common and mostly innocent — a wrong year, a transposed volume number — but a mismatch in author list or title is worth taking seriously, because it often means the citation points at a different paper from the one intended.

  • Title, matched in full rather than by its first few words
  • Author list, and specifically the first author
  • Journal name, including whether it is the journal claimed
  • Year, volume, issue and pages where applicable
  • Whether the identifier resolves at all, and to what
  • Whether the record carries notices attached to it

Corrections, retractions and expressions of concern

A paper can be real, resolvable and formally withdrawn. NLM links notices to the records they concern in both directions, so the original record carries a pointer to the notice and vice versa. Notably, the original record is not deleted: NLM retains the ability to search and retrieve a citation via the originally published form, which is why a retracted paper remains findable and citable long after the fact.

What happenedOn the noticeOn the article
A correction was publishedErratum for: …Erratum in: …
The article was retractedRetraction of: …Retraction in: …
Concern raised, not resolvedExpression of concern for: …Expression of concern in: …
Corrected and republishedCorrected and republished from: …Corrected and republished in: …
NLM does not differentiate between articles retracted because of honest error and those retracted because of misconduct. A retraction notice is therefore a signal to read further, not automatically an accusation.

Warning signs worth pausing on

None of these proves a citation is bad. Each is a reason to look harder before relying on it.

  • The DOI does not resolve, or resolves to something other than the cited work
  • The title in the citation does not match the title in the record
  • The citation points to a review or a news item while being presented as primary research
  • A secondary source is cited for a finding it is itself citing from elsewhere
  • The study population, model or conditions do not match the claim built on it
  • The abstract supports a narrower statement than the claim being made
  • A correction, expression of concern or retraction is attached to the record
  • The citation appears repeatedly across sites in identical wording, with no one quoting the source directly

The harder question

Having established that a paper exists and stands uncorrected, the remaining work is comparison. Read the claim and the relevant passage side by side and ask what has changed between them. The most common failure is not fabrication but drift: a finding in a specific model, at a specific dose, over a specific duration, restated without any of those qualifiers until it reads as a general statement. The citation is genuine. The claim it is supporting is not the finding.

Research Checklist

  • Confirm batch identity and records.
  • Document all preparation inputs.
  • Keep use within controlled laboratory workflows.
  • Do not infer applied-use suitability from guide content.

Frequently Asked Questions

If a DOI resolves, is the citation verified?

It is confirmed to exist and to point at a real object. That is the first half. It says nothing about whether the work has since been corrected or retracted, and nothing at all about whether its content supports the claim it was attached to.

What does it mean if a DOI does not resolve?

Usually a transcription error — a missing character, a truncated suffix, a stray space. Check the string against the record in PubMed or on the publisher page before concluding anything. Persistence depends on the organisations and processes behind the identifier, so a failure to resolve is a prompt to investigate rather than proof the work is fictitious.

Is a retracted paper unusable?

It is a strong signal to stop relying on it and to read the retraction notice, which states what was withdrawn and, often, why. NLM does not differentiate between retraction for honest error and retraction for misconduct, so the notice itself is where the substance is. The record remains searchable and retrievable rather than being deleted.

How can a citation be real and still be wrong?

By supporting something narrower than the claim it is attached to. A result obtained in one model, at one concentration, over one timescale can be restated without those conditions until it reads as general. Nothing about the citation is fabricated; the drift is in the sentence citing it.

Sources

The technical statements in this guide are drawn from the following. Where a definition is contested or a figure depends on method, the guide says so rather than picking one.

  1. Errata, Retractions, and Other Linked Citations in PubMedNational Library of MedicineUsed for how errata, retractions, expressions of concern and corrected-and-republished articles are linked to records in both directions, the display phrasing on notices and articles, that NLM does not differentiate retractions by cause, and that original records are retained and remain searchable.
  2. What is a DOI?DOI FoundationUsed for the definition of a DOI, conformance to ISO 26324, the prefix and suffix structure, what resolution does, that associated metadata can be updated without changing the identifier, and that persistence is a function of organisations rather than technology.