Publicationaudit.com

Publication Audit — forensic error review for any AI.

Copy the prompt below, open Claude, Gemini, or ChatGPT, paste it, and attach a manuscript (PDF). The model runs a structured forensic review: it hunts internal numerical inconsistencies, denominators that don't reconcile, headline numbers that drift across the abstract, tables, and figures, duplicate or missing references, incoherent diagnostic metrics, missing predictive values at population prevalence, extreme heterogeneity feeding a single pooled estimate, definition drift, reverse causation in screening studies, and conclusions that outrun the data — then returns an errors table, major and minor issues, and an editorial recommendation.

AI can and does make errors — including confident, fluent, professional-looking ones.

Every number, citation, quotation, recalculation, and conclusion this tool produces must be independently verified against the primary source before you rely on it, cite it, or act on it. A Publication Audit can miss real defects and can flag “errors” that are not errors. It structures a review and surfaces things to check — it does not replace the reviewer's expertise, the editor's judgment, or the author's responsibility for the manuscript. Treat every output as a list of items to confirm, never as a finding of fact, and never as medical, statistical, legal, or editorial advice. For already-published articles only: most publishers and journals treat manuscripts under peer review as confidential and advise against uploading them to external AI systems, so do not submit unpublished or under-review manuscripts here.

On references: this audit checks citation consistency and formatting — duplicates, entries cited but missing from the list, year and author errors — not whether a cited paper actually exists. A well-formed, internally consistent but fabricated (hallucinated) reference can pass clean. Confirming that a reference is real and that its DOI resolves is a separate reference-verification (RefVerify) pass against PubMed, Crossref, or the resolvable DOI, and the audit flags such references as NOT FOUND only when an actual source has been searched.

How to use it

01

Copy the prompt

Use the Copy button below. Nothing to fill in — it's ready to send.

02

Open your AI

Claude, Gemini, or ChatGPT. Use a model that accepts file uploads.

03

Paste & attach

Paste the prompt, attach the manuscript PDF, and send.

The prompt

Publication Audit — paste into any AI, then attach a manuscript

Open an AI

Claude →

claude.ai — strong on long documents and tables.

Gemini →

gemini.google.com — large context for full PDFs.

ChatGPT →

chatgpt.com — attach the PDF, then paste.

How it behaves. The prompt ends by telling the model to begin if a manuscript is attached, or to ask you for one if it isn't. The model reasons before answering. Verify before you rely on it: AI structures the review and surfaces candidate errors — every flagged number, citation, and claim should be confirmed against the primary source. This is a reviewer's aid, not a substitute for editorial judgment. What the checks are based on →

Published audits

Active Audits →

Completed Publication Audit reviews of published manuscripts, each listed in Vancouver format with its editorial recommendation. Open one to read the full errors table, major and minor issues, and the bottom line.

← Back to the tool

What the checks are based on

Publication Audit doesn't introduce new clinical thresholds — it operationalizes established reporting and quality-appraisal standards and applies them as a consistency and coherence check. For references it checks consistency and formatting, not existence; confirming that a citation is real and that its DOI resolves is a separate reference-verification pass. The audit logic draws on the following.

Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71.
Flow-diagram arithmetic and reporting completeness for systematic reviews.
Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology (MOOSE): a proposal for reporting. JAMA. 2000;283(15):2008-2012.
Reporting standard for meta-analyses of observational data.
von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. Lancet. 2007;370(9596):1453-1457.
Reporting standard for observational and registry analyses.
Reitsma JB, Glas AS, Rutjes AWS, Scholten RJPM, Bossuyt PM, Zwinderman AH. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol. 2005;58(10):982-990.
Basis for flagging separately-pooled, internally incoherent diagnostic summary points; bivariate model as the standard.
Rutter CM, Gatsonis CA. A hierarchical regression approach to meta-analysis of diagnostic test accuracy evaluations. Stat Med. 2001;20(19):2865-2884.
HSROC modeling for diagnostic-accuracy synthesis.
Whiting PF, Rutjes AWS, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155(8):529-536.
Risk-of-bias appraisal for diagnostic studies.
Wells GA, Shea B, O'Connell D, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. Ottawa Hospital Research Institute.
Quality appraisal of cohort and case-control studies; basis for scrutinizing non-discriminating uniform scores.
Youden WJ. Index for rating diagnostic tests. Cancer. 1950;3(1):32-35.
Youden index consistency check (sensitivity + specificity − 1).

← Back to the tool

Active Audits

Completed Publication Audit reviews of published manuscripts. Each entry is the manuscript's Vancouver-format citation and its editorial recommendation; open one for the full forensic review. Every flagged number, citation, and claim should be confirmed against the primary source.