Reference checking is easiest before the final deadline.
Once a draft is submitted, every citation problem becomes more expensive: a reviewer may question the source, an editor may ask for corrections, or a reader may follow a reference that does not support the sentence.
This checklist is for the last pass before submission. It is especially useful when a draft includes references from AI-assisted writing, copied bibliographies, old notes, or collaborators.
Start with LitSource Verify if you already have references to screen, or use the citation authenticity checker for a more guided workflow.
1. Confirm that each source exists
For each reference, check whether the paper, book chapter, guideline, or source record can be found through a stable identifier or reliable database.
Look for:
- DOI
- PMID
- journal title
- article title
- author list
- publication year
If a reference cannot be traced to a real source, do not keep it in the draft.
2. Match the metadata
A source can exist and still be unsafe to cite if the metadata is mismatched.
Check whether the title, authors, journal, year, volume, issue, pages, DOI, and PMID point to the same record.
Common problems include:
- DOI from a different article
- title copied with small but meaningful changes
- journal abbreviation mapped to the wrong journal
- year or page range that does not match the real record
- author order that points to another paper
Treat these as citation quality issues, not harmless formatting errors.
3. Check whether the source supports the sentence
Existence is only the first layer. The more important question is whether the cited source supports the sentence in your draft.
Ask:
- Does the source address the same claim?
- Is the population or field similar enough?
- Does the evidence point in the same direction?
- Is the sentence stronger than the paper's conclusion?
- Is the source primary evidence, a review, or only background context?
A paper can be real and still be a poor citation for the sentence.
4. Flag references generated by AI tools
AI-generated references deserve a separate pass.
Large language models can produce realistic-looking references that mix real authors, plausible journal titles, invented DOIs, and unsupported claims. Do not assume a reference is safe because it looks formal.
Use a stricter rule:
Any reference produced by an AI tool must be verified before it enters the final bibliography.
For AI-specific workflows, see the ChatGPT fake references checker.
5. Check the highest-risk citations first
If time is short, prioritize references attached to:
- central claims
- clinical or biomedical statements
- numerical claims
- claims in the abstract or conclusion
- claims a reviewer is likely to challenge
- references generated or suggested by AI
These are the citations most likely to damage trust if they fail.
6. Keep a correction trail
When you replace or remove a reference, note why.
Simple labels are enough:
- not found
- metadata mismatch
- weak support
- broader claim than source allows
- better source found
That audit trail helps if a supervisor, coauthor, or editor asks why the bibliography changed.
Final pass
Before submission, every important citation should answer three questions:
- Does the source exist?
- Does the metadata match?
- Does it support the sentence?
If you cannot answer all three, keep checking. A clean bibliography is not just a formatting task. It is part of the argument.