How to Verify a Link (DOI/Crossref/PubMed) Is Real, Not Fake: The Complete Student Guide
Quick Answer: To verify any academic link—DOI, Crossref, or PMID—follow these four steps: (1) resolve the identifier at its official registry (doi.org, crossref.org, or pubmed.ncbi.nlm.nih.gov), (2) compare the resolved title against what your citation claims, (3) search the paper title in Google Scholar or the discipline database, and (4) check Retraction Watch for retraction flags. If the resolved title doesn’t match your citation, the link is fake—even if it “works.”
If you’ve ever pasted a DOI into doi.org and seen it “resolve” without finding your paper, or noticed that a PMID search returns a completely different article, you’ve encountered one of the most common problems in academic research today: fabricated citations.
It’s not a glitch. It’s real. And it’s becoming increasingly common.
A 2026 study by Topaz et al. auditing 2.5 million biomedical papers found that 4,046 fabricated references appeared across 2,810 papers — a 12× increase from 2023. The Lancet estimated that in early 2026, 1 in every 277 biomedical papers carries at least one fabricated reference. That means you’re not just unlikely to encounter fake citations — you’re likely to cite one if you haven’t learned how to verify.
The problem gets worse when AI tools are involved. AI language models routinely hallucinate citations — generating links that look real but point to papers that don’t exist. The AI-generated citation might pass a simple “does the link resolve?” check while hiding a completely fabricated title. That’s why the four-step verification process below is the single most important skill you’ll develop for citation integrity.
In this guide, you’ll learn exactly how to verify every type of academic link — DOIs, Crossref identifiers, and PubMed PMIDs — using official registries, automated tools, and a simple workflow that catches fake citations even when they look correct. You’ll also learn about the emerging “Frankenstein citation” pattern that tricks even experienced researchers, and how to check if any paper has been retracted.
Key Takeaways
- A working DOI doesn’t guarantee the paper is real — it just means the identifier exists
- The #1 fabrication method is the “Topaz Pattern”: a real DOI paired with a fake title, which simple DOI checks miss entirely
- Frankenstein citations blend real authors and journals with fabricated titles and DOIs — they bypass surface screening
- Four-step verification catches every fabrication pattern: resolve, compare, cross-check, verify retraction
- Free automated tools (Sourcely, Scholar Sidekick) scan millions of citations for hallucinated references
Step 1: Resolve the Identifier at Its Official Registry
The first step is to take the identifier (DOI, PMID, or other link) and resolve it at the registry it belongs to. Each identifier has an official homepage — this is where the publisher or registry has the authoritative metadata for the paper.
DOIs (Digital Object Identifiers)
A DOI is the standard identifier for journal articles, books, and reports across most disciplines. It always starts with 10. followed by a number.
How to resolve a DOI:
Go to https://www.doi.org/ and paste the full DOI (including doi.org/10.xxxx/...). You can also go directly to the DOI resolution URL: https://doi.org/10.xxxx/xxxxx.
What happens next: the DOI should redirect you to the publisher’s page for the paper. The URL you land on and the title shown there is the “real” paper that this DOI belongs to.
What to look for: The page should load successfully. The paper title on the resolved page should match what your citation claims. If the title doesn’t match, you’ve hit a fabricated citation.
Example: Paste https://doi.org/10.1016/j.neuroimage.2023.120456 into a browser. If the resolved page shows “Neural network architectures for clinical imaging” but your citation claims it’s about “fMRI analysis of sleep patterns,” the DOI is being used to point to a fake paper.
PubMed PMIDs
A PMID (PubMed Central Identifier) is used exclusively for biomedical and life sciences papers indexed by the National Library of Medicine.
How to verify a PMID:
Go directly to https://pubmed.ncbi.nlm.nih.gov/?term=<PMID> — or use the URL format https://pubmed.ncbi.nlm.nih.gov/<PMID>/.
What you should see: The PubMed page for the paper, showing the title, authors, journal, and abstract.
Critical detail: PubMed has an “automatic term mapping” citation sensor (the PubMed citation sensor) that can analyze query text for citation information and return the correct citation even with minor errors or mismatches. This is particularly useful if you’ve copied a title with small typos.
Read more: PubMed User Guide — official documentation on the citation sensor, title search, and author search tools.
Crossref
Crossref is the DOI registration agency that maintains the global metadata for DOIs. You can verify a DOI’s metadata at https://www.crossref.org/.
How to check Crossref metadata:
Use the Crossref URL API: https://api.crossref.org/works/<DOI> — or use the Crossref metadata search at crossref.org.
What you should see: The bibliographic metadata registered with Crossref — title, authors, journal, publication date. This metadata should match what your citation claims.
Step 2: Compare the Resolved Title Against Your Citation Claim
Here’s where most students stop — and why they miss fake citations.
A working DOI is not proof that the paper is real. It’s only proof that the identifier exists.
The Topaz et al. study (Lancet, 2026) documented the single most common fabrication method — the “Topaz Pattern.” This pattern works like this:
A fabricated citation uses a real DOI (one that resolves successfully) but pairs it with a completely invented paper title.
The DOI resolves to a real paper — but the paper it resolves to is entirely different from what the citation claims. A simple “does the DOI work?” check misses this pattern completely.
The Topaz study showed that out of 7 fabricated references with this pattern, zero were caught by DOI resolution alone. It required a title-to-title comparison to detect.
The Topaz Pattern Explained
| Fabrication Method | What Happens | How It’s Detected |
|---|---|---|
| Ghost Paper | Entirely fake DOI — doesn’t resolve | DOI check catches it |
| Topaz Pattern | Real DOI, fake title | Title comparison catches it |
| Frankenstein Citation | Real DOI, real author, fake title | Title comparison catches it |
| Phantom Journal | DOI resolves to a non-existent journal | Title comparison + journal name check |
| Mismatched IDs | DOI and PMID resolve to different papers | Cross-check catches it |
The four-step workflow catches every fabrication pattern because it forces you to verify the title, not just the identifier.
The Frankenstein Citation Problem
In early 2026, Grounded AI co-founder Joe Shockman coined the term “Frankenstein citations” to describe a particularly sneaky fabrication pattern: a citation that blends real elements with fabricated ones.
A Frankenstein citation might use:
- A real author name (e.g., a real researcher with real publications)
- A real journal title (e.g., a legitimate academic journal)
- A fabricated paper title (the invented title of a paper that never existed)
- A fabricated DOI (a DOI that resolves, but to a different paper)
These citations bypass surface-level human screening because they “look” correct at a glance — the author name exists, the journal name exists, the DOI resolves. But the title is entirely fabricated.
The only way to catch a Frankenstein citation is to compare the title you’re citing against the title on the resolved DOI page. This is what makes the four-step verification process mandatory.
Step 3: Cross-Check the Title in Scholar Databases
The second step is to take the resolved title (from the DOI resolution page) and search it in discipline-specific databases.
Google Scholar
Search the full paper title in https://scholar.google.com/. If the paper exists, it should appear in Google Scholar with a snippet showing the title, authors, and journal.
What this catches: If the resolved title doesn’t appear in Google Scholar at all, you’re likely looking at a fabricated citation. If the resolved title appears but in a different journal or under different authors, you’ve caught a mismatch.
PubMed (for biomedical papers)
Search the paper title at https://pubmed.ncbi.nlm.nih.gov/. Use PubMed’s advanced search to query by title, author, and journal name.
What this catches: PubMed indexes over 35 million citations from journals worldwide. If a paper claims to be in a biomedical journal and it’s not indexed in PubMed (assuming it’s actually published in a biomedical journal), the citation is fabricated.
Discipline-Specific Databases
| Discipline | Primary Database | URL |
|---|---|---|
| Computer Science | arXiv | https://arxiv.org/ |
| Engineering | IEEE Xplore | https://ieeexplore.ieee.org/ |
| General Academic | Scopus | https://www.scopus.com/ |
| General Academic | Web of Science | https://www.webofscience.com/ |
| Physical Sciences | ADS (NASA) | https://ui.adsabs.harvard.edu/ |
If the resolved title appears in these databases, the paper is likely real. If it doesn’t, the citation is suspicious.
Step 4: Check Retraction Status
The final verification step is checking if the paper has been retracted. A paper might be real (and not fabricated), but still retracted for errors, misconduct, or other reasons.
Retraction Watch Database
Go to https://www.retractionwatch.com/ and search the paper by DOI, PMID, title, or author. The Retraction Watch Database tracks retractions across all disciplines.
What this catches: If a paper you’re citing has been retracted, you should not use it as a source. This is critical for citation integrity — citing a retracted paper is academically dishonest, even if the paper was not fabricated.
Crossref Retraction Metadata
Use the Crossref URL API to check for retraction status: https://api.crossref.org/works/<DOI>/retracted — if the DOI is retracted, Crossref will return retraction metadata.
PubMed Retracted Publication Flag
PubMed marks retracted papers with a “Retracted” flag in the PubMed record. The NCBI publication data also tracks correction policies, errata, and duplicate publications.
Read more: PubMed Help for Data Providers — official correction policy and PMID assignment process; explains how NLM handles retractions.
Using Automated Verification Tools
Manual verification is thorough, but there are also automated tools you can use to speed up the process.
Sourcely (Citation Verification & AI Citation Checker)
Sourcely’s free citation checker (https://www.sourcely.net/citation-verification) scans your citations against a 200M+ paper database and returns a verdict: Matched, Mismatch, Ambiguous, or Not Found.
Read more: Sourcely Citation Verification — step-by-step guide on catching AI-hallucinated citations
Scholar Sidekick (Citation Integrity & Verifier)
Scholar Sidekick’s Citation Verifier (https://scholar-sidekick.com/tools/citation-verifier) and Identifier Validator (https://scholar-sidekick.com/tools/identifier-validator) document the Topaz pattern and provide precision/recall data: 87.3% fabrication recall, 0.87% false-accusation rate.
Read more: Scholar Sidekick Citation Verifier — technical documentation of the Topaz et al. fabrication pattern
Read more: Scholar Sidekick Citation Integrity — measured precision/recall data
The Four-Step Verification Checklist (Your New Workflow)
Here’s your practical verification workflow. Keep this checklist open when you’re building your bibliography:
| Step | Action | Tool | Verdict |
|---|---|---|---|
| 1. Resolve | Paste DOI/PMID at official registry | doi.org, pubmed.ncbi.nlm.nih.gov | Does it resolve? |
| 2. Compare | Check resolved title vs citation title | Browser comparison | Titles match? |
| 3. Cross-check | Search title in discipline database | Google Scholar, PubMed | Paper indexed? |
| 4. Verify | Check retraction status | Retraction Watch, Crossref | Retraction flag? |
If any step fails: the citation is fabricated or retracted. Do not cite it.
Why This Matters for Your Grades
Citing fabricated citations is one of the most common reasons papers are flagged for academic misconduct. Professors and journal editors increasingly use automated verification tools to check references. If you cite a fabricated paper, your work may be flagged — even if you didn’t know it was fake.
The Topaz study documented that 98.4% of papers with fabricated references received no publisher correction. The fabricated references simply stay in the literature, misleading future research.
That’s why verification is not optional. It’s the minimum standard for academic writing.
What To Do Next
Once you’ve verified your citations, you’ll know that every link you include is real, resolvable, and accurate. You’ve done the work that most students skip — and that’s what separates strong academic writing from weak work.
If you need help with academic writing, editing, or citation formatting, we can assist. Explore our writing and editing services to get started.
Related Guides
- How to Cite AI Chat Conversations in Academic Papers — APA, MLA, Chicago citation formats
- APA Citation 2026: New Rules for AI Sources — Updated AI citation rules across disciplines
- Citation Management Tools: Zotero vs Mendeley vs EndNote — Tools for organizing verified citations
- Citation Styles Compared: APA vs MLA vs Chicago — Field-by-field formatting guidance
FAQ
Does a working DOI guarantee the paper is real? No. The “Topaz Pattern” shows that a real DOI can point to a different paper than what your citation claims. Always compare the resolved title against your citation title.
What are Frankenstein citations? Frankenstein citations are hybrid hallucinations that use real author names and journal titles with fabricated paper titles and DOIs. They bypass surface-level human screening because they “look” correct at a glance.
How do I check if a paper has been retracted? Search the Retraction Watch Database by DOI, PMID, title, or author. Check Crossref metadata at https://www.crossref.org/. PubMed flags retracted papers with a “Retracted” status.
Is there an automated tool to check citations? Yes. Sourcely and Scholar Sidekick offer free citation verification tools that scan millions of papers for hallucinated references.
How many fabricated citations are there in literature? The Topaz et al. Lancet 2026 study audited 2.5M biomedical papers and found 4,046 fabricated references across 2,810 papers — a 12× increase from 2023 to 2026.
