Check numeric consistency across manuscript sections and tables

Spot-check sample counts, p-values, percentages, and statistical estimates across different sections of an academic draft.

You getDiscrepancy table with columns, clean values confirmation list and critical flags requiring co-author data re-check

Prompt

Prompt
Act as an expert academic copyeditor and data auditor. Cross-check the provided manuscript sections to find numerical discrepancies, mathematical mismatches, and reporting contradictions.
- Cross-reference sample counts (n), percentages, degrees of freedom, effect sizes, test statistics, and p-values between text, tables, and summary statements.
- Identify rounding errors, percentages that do not sum to 100%, and mismatched confidence intervals.
- Point out instances where abstract values differ from the results section or tables.
- Highlight missing denominator context or ambiguous statistical units.
Output format:
1. Discrepancy table with columns: Value Location A, Value Location B, Discrepancy Found, Probable Cause / Suggested Fix
2. Clean values confirmation list (metrics verified as consistent)
3. Critical flags requiring co-author data re-check

Here are the manuscript excerpts:

Example output

Example for: an epidemiological paper draft with mismatched cohort counts between Abstract and Table 2

1. Discrepancy Table

| Value Location A | Value Location B | Discrepancy Found | Probable Cause / Suggested Fix |
| :--- | :--- | :--- | :--- | :--- |
| Abstract: "n = 412 patients enrolled" | Table 1 Total: "N = 418" | Difference of 6 participants between abstract and baseline demographic table | Abstract likely reflects final evaluable set rather than enrolled total. Clarify: "418 enrolled, 412 completed protocol". |
| Results (Para 2): "34.2% (n = 141)" | Table 2 (Row 1): "n = 141 (33.7%)" | Percentage mismatch for the same numerator | Results text divides by 412 (141/412 = 34.2%), while Table 2 divides by 418 (141/418 = 33.7%). Unify denominator across both. |
| Discussion: "p < 0.001" | Table 3: "p = 0.042" | Abstract and Discussion claim high significance for secondary outcome, contradicting Table 3 | Secondary biomarker analysis was reported as primary outcome metric. Update Discussion to state exact p = 0.042. |

2. Clean Values Confirmation List
  • Primary outcome mean age: 58.4 years (consistent across Abstract, Table 1, and Results).
  • Gender breakdown: 210 female, 202 male in evaluable population matches across text and figures.
  • Odds ratio for treatment arm A: OR 1.84, 95% CI [1.12, 3.01] matches between Results and Discussion.
3. Critical Flags Requiring Co-Author Data Re-check
  • Re-confirm whether Table 2 should report percentages based on enrolled participants (418) or completed participants (412).
  • Check whether the secondary endpoint p-value in Table 3 was adjusted for multiple comparisons.
Written by AI from this prompt. Your results will vary.

Paste your Abstract, Results excerpt, and related Data Tables directly after the prompt into ChatGPT, Claude, or Gemini. As a follow-up, ask the model to produce corrected sentences that match your authoritative source table.

Make it repeatable

Keep using this prompt without pasting it in every time.

Keep exploring

More in Academic Writing & Study
Academic Writing & Study

Check paper claims against study evidence

Flag overstatements, unwarranted causal claims, and speculative generalisations in draft academic text. Ideal before journal submission.

Academic Writing & Study

Draft a defensible study limitations section

Draft a balanced academic limitations section that transparently acknowledges methodology constraints while safeguarding your study's original contribution.