Do Journals Check References in 2026? Trends, Real-World Data, and What Authors Must Do

TL;DR — What You Need to Know First

• Most peer reviewers do not systematically verify every reference in a manuscript (76.7% skip thorough checking)
• 98.4% of papers with fabricated references received no publisher action after the 2026 Lancet audit
• The editorial desk is where your manuscript will face its first real test — 70-96% of submissions are desk-rejected before peer review begins
• Automated reference-checking tools (scite.ai, CrossRef, Paperpal) are now mainstream at major publishers
• Your pre-submission checklist is your safety net — here’s exactly what to do before you hit “submit”

The 2026 Wake-Up Call: 12-Fold Surge in Fabricated Citations

In May 2026, a groundbreaking audit published in The Lancet laid bare a problem that’s been quietly growing for years. Topaz and colleagues audited 2.5 million biomedical papers and found 4,046 fabricated citations — references that don’t exist, have incorrect authors, or point to completely unrelated papers.

The scale is staggering. In early 2026, 1 in every 277 papers contained at least one fabricated reference. Compare that to 2023, when the rate was 1 in 2,828. That’s a 12-fold increase in just three years.

What’s more, review articles had a 57% higher fabrication rate than standard research papers — the exact type of paper students, early-career researchers, and anyone doing a literature review would turn to for background.

And here’s the uncomfortable truth: 98.4% of affected papers received no publisher action. Journals didn’t retract, didn’t issue errata, and didn’t even flag the papers.

This is not a hypothetical problem. AI-generated text with hallucinated citations is real, widespread, and the journals you’re submitting to are only just starting to detect it.

What Actually Happens During Journal Screening?

When you submit a manuscript, two stages stand between you and peer review — and both involve reference checking.

The Editorial Desk (Where 70-96% of Submissions Die)

Before a single reviewer ever sees your paper, an editor performs a desk screening. This is not a formality. 70-96% of manuscripts at elite journals are rejected at this stage.

The triggers are mechanical and predictable:

• Broken or unresolvable DOI links
• Missing reference list items
• Formatting inconsistencies across citation styles
• Suspiciously recent or niche sources in fields where they’re unlikely to be cited

If your references don’t survive this pass, peer review never begins. You don’t get feedback on methodology, data, or conclusions — your paper simply doesn’t make it to that stage.

What Peer Reviewers Actually Check

Once your paper lands on a reviewer’s desk, the reference checking gets selective.

A survey of 203 researchers at CHI and AAAI 2026 conferences revealed that reviewers approach citation checking in widely different ways — and many do so lightly. An AI Overview consensus from multiple academic communities reports that 76.7% of reviewers do not thoroughly verify every reference, and 80% have never suspected a fake or hallucinated citation in a paper they reviewed.

Why? Because reviewers are stretched thin. Their primary obligations are to:

• Evaluate whether the methodology is sound
• Assess whether the data supports the conclusions
• Judge novelty and relevance to the field

Reference formatting and verification are typically delegated to copy-editors. Reviewers read references for context but don’t resolve DOIs or check every citation against its source.

Bottom line: You cannot rely on reviewers to catch broken references. The responsibility falls on you.

What’s Changed in 2025-2026

Two shifts are reshaping how journals handle references — and neither happened before 2025.

Automated Reference Checking Goes Mainstream

Journals are no longer relying on human reviewers to catch hallucinated citations. Automated tools are now embedded in editorial workflows:

• scite.ai — Highlights “contextual verification” by checking whether a citation is supported by the source document
• PeerReviewAI — Analyzes manuscript structure, reference quality, and potential citation manipulation
• Paperpal — Combines writing assistance with reference validation checks
• CrossRef — Real-time DOI resolution integrated into editorial management systems (ScholarOne, Editorial Manager)
• Citely.ai — Automated bibliography generation with built-in reference verification

A 2026 AI Overview consensus across academic communities found that 70.2% of researchers support automated reference checks, recognizing that human reviewers cannot realistically verify every citation at scale.

COPE Updates on Citation Integrity

The Committee on Publication Ethics (COPE) has issued updated guidance in 2025-2026 that explicitly addresses:

• AI-hallucinated citations as research integrity threats
• Paper mill-generated references in synthetic manuscripts
• The reviewer obligation to refrain from citation manipulation

These aren’t theoretical guidelines. Journals are incorporating COPE language into their author guidelines and reviewer checklists. If you submit a paper with fabricated references, the editor’s automated screening tool flags it — and the editor now has formal guidance on how to handle it.

Your Pre-Submission Reference Checklist

This is the actionable part. Here’s exactly what you need to do before you submit a manuscript — a step-by-step workflow that protects you from desk rejection and accidental citation errors.

Let’s walk through each step.

Step 1: Resolve Every DOI

Copy the DOI string from your reference list and paste it into https://doi.org. For example, if your reference says doi:10.1016/S0140-6736(26)00001-X, paste the full DOI string 10.1016/S0140-6736(26)00001-X into the resolution tool.

If the DOI resolves to a page that doesn’t match your citation (wrong title, wrong authors, or a completely unrelated paper), your reference is hallucinated or incorrectly attributed. This is the most common error we see — especially with AI-generated text.

Step 2: Search Crossref

Crossref is the DOI registration authority. Searching by exact title on https://www.crossref.org confirms that the paper actually exists and that your bibliographic details are accurate.

Pay attention to year mismatches. A paper cited with “2023” when the actual publication year is 2024 is a common AI hallucination pattern.

Step 3: PubMed Check (for Biomedical Fields)

If you’re citing biomedical, clinical, or life-science papers, resolve each citation against the PubMed database. This confirms:

• The PMID (PubMed ID) resolves to the correct paper
• The paper has not been retracted
• The abstract and methods match what you’re citing

Step 4: Check the Retraction Watch Database

If a paper you cited has been retracted, you need to know why. A retracted paper cited in your manuscript creates a cascade problem: reviewers will flag it immediately, and editors can escalate to research misconduct if the retraction is severe.

Check Retraction Watch or the Retraction Watch Database for every paper you’re citing. It’s a free resource — and the single most important quality control step an author can perform.

Step 5: Run an Automated Reference Check

Tools like scite.ai, Paperpal, and PeerReviewAI now offer reference validation features:

• scite.ai flags citations where the source doesn’t actually support the claim
• Paperpal checks bibliography formatting and consistency
• PeerReviewAI flags suspicious patterns that editors use to detect citation manipulation

Even free tools like Zotero’s reference manager integration can catch broken links and formatting inconsistencies.

Step 6: Orphan Check

The final manual check: go through your manuscript and for every in-text citation (Smith 2024), verify that there’s a corresponding entry in your reference list. If you cite a source that doesn’t have a reference entry — or a reference entry that has no in-text citation — that’s an orphan and a formatting error that editors will flag.

Extra Links You Forgot About: Body-Text Hyperlinks and Supplementary Materials

The six-step checklist above covers your reference list — the formal citations at the end of your paper. But there’s another layer of links that editors check, and which researchers increasingly omit or include without verifying:

Body-Text Hyperlinks (Not Citations)

Many modern manuscripts include URLs directly in the text — not as formal citations, but as inline links to figures, datasets, code repositories, or supplementary materials. These are particularly common in:

• Computer science and AI papers (GitHub repos, Hugging Face model cards, arXiv source code)
• Open-science and reproducible-research manuscripts (OSF projects, Zenodo datasets, supplementary appendices)
• Clinical trials (ClinicalTrials.gov registration numbers, open-data repositories)
• Data-science and AI papers (live dashboards, model endpoints, API documentation)

Unlike formal citations in your reference list, inline URLs get checked differently. Editors don’t resolve DOIs for these — they click them. A broken link in the body text is the easiest way to fail a desk review.

Supplementary Materials

Modern publishing frequently requires supplementary materials — appendices, raw data files, additional figures. Many journals embed these as URLs in the manuscript body rather than as part of the reference list.

The mistake students and early-career researchers make: they link a supplementary file that they’ve uploaded to their own cloud drive, which then expires or becomes private before publication. Always use a permanent, citable URL:

• OSF (osf.io) — free, assigns a permanent DOI to datasets
• Zenodo — free, CITES-certified permanent archiving
• GitHub — free, versioned, permanent for code
• Figshare — free for researchers, assigns DOIs

If you link a personal Google Drive or Dropbox folder, the editor will click it — and if access expires, your manuscript looks careless.

The “Link vs. Identity” Distinction

This concept was documented by Paperpal in their 2026 reference-checking guidance and confirmed by independent research. A working URL satisfies a link check — it loads in a browser. But it fails an identity check if the URL resolves to the wrong paper.

For example, a DOI link might resolve to a paper’s landing page (the “link” is valid), but the actual PDF behind the DOI has been swapped or redirected. Or the title of the paper on the resolved page might not match what you cited. This distinction matters most when you’re citing unfamiliar papers or AI-generated text.

When to Check Your Links: Timing Your Verification Pass

Knowing what to check is only half the equation. Timing matters — and most researchers check too late.

The most common mistake: Researchers spend weeks perfecting their methodology and analysis, then rush through references the night before submission. That’s when hallucinated AI citations surface — and when supplementary material links break. Check references as you cite them, then run a full verification pass after formatting.

The golden rule: Your verification should happen in the journal’s template, not in your draft. If you format your manuscript first and then check links, the formatting process may alter URLs, embed PDFs differently, or change how supplementary materials are embedded. Always verify in the final version.

When Bad References Slip Through: Consequences

What happens if a journal or reviewer discovers fabricated or incorrect references? The consequences cascade.

At the Desk-Review Stage

• Desk rejection (most likely) — Your paper never reaches peer review
• Request for revision — The editor gives you a chance to fix the references and resubmit

After Peer Review and Acceptance

• Corrigendum or erratum — A formal correction notice published alongside your paper
• Retraction — The paper is withdrawn; this is your professional record’s permanent notation

At the Student and Early-Career Level

• For students, fabricated references in a thesis or dissertation can delay graduation
• For early-career researchers, a retraction in the early stages of a career carries disproportionate professional consequences
• Systematic reviewers who cite retracted papers risk their entire review being called into question

This is not an alarmist warning. The statistics we cited earlier (1 in 277 papers with fabricated references, 98.4% receiving no publisher action) show that most bad references are honest AI errors — not misconduct. But the onus is on you to verify.

Bottom Line for Authors

The landscape has shifted. In 2026, you’re not just responsible for writing a strong paper — you’re responsible for the integrity of every single citation in it.

The editorial desk rejects 70-96% of submissions for format and reference issues alone. Peer reviewers, by design, do not thoroughly verify every reference. Automated tools exist and are being deployed at publishers — but they catch what humans miss, and sometimes miss what automated screening misses.

Your checklist is your safety net. Run the six-step process above. Resolve every DOI. Search Crossref. Check retraction status. Run an automated pass. Check for orphans. Then click every body-text link, supplementary URL, and GitHub repo. Verify everything in the journal’s template.

Then submit with confidence.

Related Guides

• How to Verify a Link (DOI/Crossref/PubMed) Is Real — Step-by-step DOI and Crossref verification
• Frankenstein Citations: How to Spot Partially Incorrect AI References — Detection checklist for partially fabricated references
• How to Cite Social Media and Instagram Posts: APA, MLA, Chicago Guide 2026 — Citation formatting for non-traditional sources
• Fake LLM Citations: Why AI Fabricates References (Statistics 2023-2026) — Understanding AI citation failure modes
• How to Check for Retracted Articles (Retraction Watch, Alerts) — Complete retraction-checking workflow