You hit Submit at 11:47 PM. The confirmation email arrives at 11:48. You feel the brief window of relief that comes before the first doubt.
Did you actually check citation 14?
It is not a question of whether the DOI resolves. You checked that. It is whether the paper says what you claimed it says. You skimmed the abstract three weeks ago when you dropped the citation into the draft. You did not read the conclusion. You are not sure the direction of the finding matches the direction of your claim.
The journal’s peer review process will check. The question is whether you want them to be the first ones looking.
The 6-Hour Checklist
Each check catches a specific failure mode. Each one maps to something The Gate in the Trust Stack catches automatically.
You do not need all six hours. You need the six checks. Here they are, in order of importance.
Check 1: Claim-Source Alignment (90 minutes)
For each citation attached to a central claim, open the cited paper. Read the abstract. Read the conclusion. Does the paper’s directional finding match the direction of your claim?
This is the check that catches backwards citations, real papers, real DOIs, inverted claims. In production audits, backwards citations appear in roughly 7% of AI-generated citation sets that passed DOI verification alone. For a 40-citation paper, that is 2-3 misaligned references waiting for Reviewer #2.
What to flag: any citation where the paper’s finding is null, contested, or directionally opposite to your claim.
Check 2: DOI Existence (20 minutes)
Run every DOI through CrossRef. Does it resolve? Does the metadata (title, authors, year) match what your bibliography says?
This catches fabricated citations and hallucinated identifiers, the failure mode that got 156 lawyers sanctioned and that Nature estimated appeared in 110,000+ publications in 2025 alone. It catches roughly 40% of citation failure modes based on our production verification data.
What to flag: any DOI that does not resolve or resolves to a different paper than your bibliography describes.
Check 3: Polarity (30 minutes)
For each paper cited as settled evidence, check whether the broader literature supports or contradicts it. Scite.ai classifies 1.6 billion citation statements as supporting, contradicting, or mentioning. A paper with a high contradiction ratio is contested evidence, and presenting it as settled is a risk.
What to flag: any paper where contradicting citations exceed 20% of total citation statements. This is the check that catches papers the field disputes, not papers that do not exist.
Check 4: Internal Consistency (30 minutes)
Cross-reference your citations. Does citation 7 make a claim that citation 23 contradicts? Does your literature review acknowledge a limitation that your analysis section ignores?
Reviewers catch these because they read the paper end to end. You need to read your own paper the way a skeptical reviewer would, looking for seams between sections.
What to flag: any pair of citations that make conflicting claims, and any section that ignores a limitation acknowledged elsewhere.
Check 5: Recency and Relevance (20 minutes)
Are you citing a 2018 paper for a claim that has been superseded by 2024 research? Are you citing a preprint when a peer-reviewed version now exists?
This is not about having the newest citations. It is about not having the wrong ones, outdated studies that no longer represent the field’s current understanding.
What to flag: any citation to a paper that has been superseded, retracted, or significantly updated since you last checked.
Check 6: Scope Boundaries (10 minutes)
Does each citation support the specific claim it is attached to, or does it support a broader claim than you are making? A paper about college students does not support a claim about “people.” A mouse study does not support a claim about clinical outcomes.
This is the check that catches overreach, citing a narrow study for a broad claim. Reviewers notice this because it is one of the easiest mismatches to spot.
What to flag: any citation where the study population, methodology, or scope is narrower than the claim it supports.
The Gap Between This Checklist and Systematic Verification
You can run all six checks by hand. The question is whether you want to.
The manual checklist catches the same failure modes that The Gate catches automatically. The difference is not what gets caught. It is:
- Speed: 2-6 hours by hand vs. minutes through systematic verification
- Completeness: A tired researcher at 11 PM skips checks 3 and 6. The system does not skip.
- Audit trail: The manual check leaves no record. Systematic verification produces a structured log, what was checked, what was found, what was flagged, that you can produce if a reviewer, editor, or compliance officer asks.
- Repeatability: When you revise the paper in three months, you run the checklist again from scratch. The system re-runs only the changed citations.
The checklist is useful as a standalone tool. Run it. It will catch things. But it is also the blueprint for what systematic verification does automatically, which is why the researcher who runs the checklist and the researcher who runs The Gate both submit papers that survive peer review. The second one does it in less time, with more confidence, and with a record to prove it.
A 20-Minute Citation Check Before Submission Saves a 3-Month Revision Cycle After
If you do not have time for all six checks, run these three:
- Claim-source alignment for every citation attached to a central claim (the ones your paper’s argument depends on)
- DOI existence for every citation (the 20-minute sweep that catches fabrication)
- Polarity for the 5 most important citations (the ones that carry the most weight in your argument)
These three catches cover the failure modes that actually trigger resubmission: backwards citations on central claims, fabricated references, and contested evidence presented as settled.
The other three checks matter. But if you have 20 minutes and a deadline, these are the ones that prevent the letter from Reviewer #2.
If you wish you had a system that runs these checks automatically before every submission instead of a checklist you run manually before some of them, we can help you build it. Request a research intake.