Skip to main content
Systematically check whether the claims in a medical communications document are accurately supported by the cited references — identifying unsupported claims, numerical errors, overstatements, and missing qualifiers before the document reaches formal review.
Risk tier: High  ·  Review requirement: Expert review — verification results must be assessed by a qualified medical writer or reviewerThis workflow supports your review — it does not replace it. Automated tools do not catch every error. A verified flag means the tool found a potential issue; a clean scan does not guarantee accuracy.

​
What AI does

  • Scans large documents with many claim-reference pairs rapidly
  • Identifies numerical mismatches (wrong p-values, incorrect percentages, swapped comparator arms)
  • Flags claims that use stronger language than the source supports (e.g., “significantly improved” where the source reports a trend)
  • Structures the verification output into a reviewable format

​
What AI cannot do

AI checks whether a claim is technically supported by the cited reference. It does not check whether the claim is used in an appropriate context, whether important information from the reference has been omitted, or whether the reference cited is the best available evidence. These require human judgement.

​
Before you start

  • Ensure every verifiable claim in the document has a clear reference citation. Number or label claims if the document does not already do so.
  • Obtain the full text of every cited reference. Partial verification against abstracts only creates a false sense of security — do not rely on abstracts alone.
  • Confirm that references are finalised. Running this workflow on a document with placeholder references wastes effort.
If you do not have access to the full text of all cited references, do not proceed. Partial verification is not verification.

​
Steps

1

Prepare the document

Ensure every verifiable claim has a clear reference citation. If the document does not already map claims to references, label them before starting.
2

Gather all reference materials

Obtain the full text of every cited reference. Use publisher PDFs wherever possible — scanned PDFs may have OCR errors in results tables.
3

Run automated verification with RefCheckr

Submit the document and reference materials to RefCheckr for a first-pass claim-to-reference comparison.
4

Review every flagged item

For each flag, open the cited reference and manually assess whether the claim is supported, partially supported, or not supported. Do not dismiss flags without checking.
5

Spot-check unflagged items

Automated tools do not catch everything. Manually spot-check 20–30% of unflagged claims, prioritising those with specific numerical data, comparative language, or endpoint results.
6

Assess context for supported claims

A claim may be technically supported by a reference but used in a context that changes its meaning. Review each claim in the context of the full document narrative, not just as an isolated claim-reference pair.
7

Correct and annotate

Fix inaccurate claims, update references, and document the verification outcome for each item reviewed.
8

Re-verify after significant changes

If you made substantial corrections, run the verification again to confirm the changes resolve the flagged issues without introducing new ones.

​
Prompt pattern

Use this pattern when running manual verification or using a general-purpose LLM to support the process. 
You are a medical writing QC assistant. Your task is to verify the following claims against their cited references.

For each claim:
1. State the claim exactly as it appears in the document
2. Identify the cited reference
3. Find the relevant section of the reference that should support the claim
4. Assess whether the reference supports the claim: SUPPORTED / PARTIALLY SUPPORTED / NOT SUPPORTED / CANNOT VERIFY
5. If partially supported or not supported, explain the discrepancy
6. Flag any numerical mismatches (different p-values, percentages, sample sizes, etc.)
7. Flag any claims where the document uses stronger language than the reference

Document claims:
[INSERT DOCUMENT TEXT OR CLAIM LIST WITH REFERENCE NUMBERS]

Reference materials:
[INSERT REFERENCE TEXTS — one per reference, clearly labelled]

Rules:
- Compare claims strictly against the cited reference. Do not use general knowledge to fill gaps.
- If a reference does not contain information relevant to the claim, mark it as NOT SUPPORTED by this reference.
- Note any qualifiers in the reference that are missing from the claim (e.g., subgroup, post-hoc, exploratory).

​
Interpreting verification results

The reference directly supports the claim as written. Spot-check the context — a supported claim used in a misleading context is still a problem.

​
Human review checklist

  • All flagged claims have been manually reviewed against the cited reference
  • A sample of unflagged claims has been spot-checked for accuracy
  • Numerical data (p-values, CIs, percentages, sample sizes) verified against source
  • Claims use language consistent with the strength of evidence (no overstatement)
  • Qualifiers in the source are preserved in the claims (subgroup, post-hoc, exploratory)
  • Safety claims are verified with the same rigour as efficacy claims
  • All NOT SUPPORTED or PARTIALLY SUPPORTED flags have been resolved
  • Changes made during verification are documented
  • The document is ready for formal review with reference accuracy confirmed

​
Common failure modes

RiskWhat to look for
False negativesAutomated tools miss numerical errors when the reference and claim discuss the same metric — manually verify all specific data points in key claims
Context blindnessA claim is technically supported but used in a comparative context the reference does not support — assess claims in the context of the full document
Reference text qualityOCR errors in scanned PDFs cause tools to misread numbers — use publisher PDFs where possible
Scope limitationVerification confirms that claims match cited references; it does not assess whether those references are the best available evidence

​
Relevant tools

RefCheckr

Primary tool for systematic claim-to-reference comparison.

​
Next steps

Check Compliance

Pre-screen for compliance signals after verifying references.

Final Review

Complete the QC process before submission or publication.