Hepatology's AI Policy: AASLD and Wolters Kluwer Rules for Liver Disease Authors

Non-invasive fibrosis assessment and AI

The hepatology field has been moving toward non-invasive assessment of liver fibrosis for years. Elastography interpretation, serum biomarker panels (FIB-4, APRI, ELF), and AI-assisted ultrasound analysis are all active areas of research published in Hepatology. When your paper is about these AI diagnostic tools, the AI is your research subject.

The distinction between research AI and writing AI remains critical. If you developed a deep learning model for automated elastography interpretation and also used ChatGPT to polish your manuscript, keep the descriptions completely separate:

• Research AI: standard Methods section with architecture details, training/validation data, performance metrics
• Writing AI: dedicated disclosure paragraph naming the tool and its purpose

NASH/MASH clinical trials

Hepatology publishes practice-changing clinical trials for metabolic dysfunction-associated steatohepatitis (MASH, formerly NASH). Resmetirom, other FXR agonists, thyroid hormone receptor-beta agonists, these trials have transformed the treatment landscape. The papers reporting their results carry enormous clinical weight.

For trial reports, AI can help with language. It shouldn't help with interpreting histologic response rates, characterizing adverse event profiles, or drawing conclusions about clinical benefit. Hepatology's reviewers include hepatologists who run these trials and serve on DSMB committees. They'll notice if the language in your efficacy section has the hedged, overly balanced quality that AI tools tend to produce when discussing clinical outcomes.

Liver transplant data

Hepatology publishes substantial transplant research: allocation algorithms, donor-recipient matching studies, post-transplant outcomes. These papers often involve large registry datasets (UNOS, Eurotransplant) with sensitive patient information.

Never process transplant registry data through external AI tools like ChatGPT. Even de-identified data can create compliance issues if it's uploaded to a cloud-based AI service. This isn't just an AI policy concern, it's a data governance issue that your IRB and registry access agreement would also prohibit.

AASLD practice guidelines

Hepatology is the primary publication venue for AASLD Practice Guidelines. These cover hepatitis B management, hepatitis C treatment, liver cancer surveillance, alcohol-related liver disease, MASH management, and more. Millions of patients worldwide are treated according to these documents.

The AI stakes for guideline papers are at their highest. While there's no separate formal rule for guidelines versus original research, the editorial expectation is clear: guideline recommendations must be entirely the product of the expert panel's deliberation. AI involvement in drafting or formulating guideline statements would undermine the entire framework of evidence-based expert consensus that these documents represent.

For an original research article (e.g., MASH biomarker study):

"During the preparation of this manuscript, the authors used ChatGPT (GPT-4o, OpenAI) to improve the clarity of the Introduction and Discussion sections. All AI-generated suggestions were reviewed and edited by the corresponding author (A.B.) and the senior hepatologist (C.D.). No AI tools were used in study design, sample analysis, biomarker validation, or clinical interpretation. The authors take full responsibility for the content of this article."

For a clinical trial report (e.g., Phase III MASH trial):

"The authors used Claude (Claude 3.5, Anthropic) to improve the English language of the Methods section. No AI tools were used for trial design, statistical analysis, histologic assessment, efficacy evaluation, safety reporting, or interpretation of clinical outcomes. All clinical conclusions were drawn by the study investigators based on the pre-specified statistical analysis plan reviewed by the Data Safety Monitoring Board. The authors take full responsibility for the published content."

For a transplant outcomes study:

"ChatGPT (GPT-4, OpenAI) was used to improve the readability of the Results and Discussion sections. No patient-level data from the transplant registry was processed through any external AI tool. Statistical analyses were performed using SAS 9.4 (SAS Institute) and R (v4.3.2) by the study biostatisticians. The authors take full responsibility for the published content."

For a systematic review:

"During the preparation of this systematic review, the authors used Claude (Claude 3.5, Anthropic) to improve sentence-level readability of the Discussion section. The literature search, study screening, data extraction, quality assessment, and evidence synthesis were performed entirely by the author team. No AI tools assisted with any stage of the systematic review process. The authors take full responsibility for the published content."

Each disclosure follows the same pattern: name the tool, describe the scope, clarify what wasn't AI-assisted, and assert responsibility. For clinical papers, the explicit disclaimer about clinical interpretation is essential. Reviewers and editors look for it.

The consequences of undisclosed AI use at Hepatology follow COPE guidelines, implemented through both AASLD and Wolters Kluwer infrastructure:

During peer review. Reviewers increasingly comment on prose that reads like AI output. If a reviewer flags this concern, the editor will ask for clarification. This doesn't automatically doom your paper, but it shifts the editorial dynamic. You're now defending your integrity instead of discussing your science.

After acceptance, before publication. The paper can be held in production while the issue is resolved. You'll need to provide a disclosure statement and explain why it wasn't included originally. This delays your publication and creates an editorial record.

After publication, the consequences escalate:

1. Correction. If AI use was limited to language editing, a published erratum adding the disclosure may be sufficient. Your paper gets a correction notice that's permanently linked to it in PubMed and other databases.

1. Expression of concern. If AI involvement raises doubts about the scientific content, say, AI was used to draft clinical interpretations for a fibrosis study, the editor may issue a formal expression of concern. This is a public statement that something about the paper is under investigation.

1. Retraction. For cases where AI use was extensive enough to undermine confidence in the findings, retraction is possible. In hepatology, where clinical guidelines are built on published evidence, a retracted paper can have downstream effects on treatment recommendations.

1. Institutional notification. In serious cases, the journal notifies the authors' institution. For faculty at academic medical centers, this triggers a formal research integrity inquiry. These investigations typically take 6 to 12 months and can affect promotion, grant funding, and clinical privileges.

1. Cross-journal pattern detection. Wolters Kluwer's infrastructure allows them to identify patterns across their portfolio. If the same author has undisclosed AI use flagged at multiple Wolters Kluwer journals, the consequences compound.

Key observations:

Hepatology and Journal of Hepatology are the two primary competitors in liver disease research. Both follow the same basic AI rules but draw from different societies: AASLD (North American) and EASL (European). The policy substance is nearly identical. The main practical differences are in submission systems, Hepatology uses Wolters Kluwer's Editorial Manager while Journal of Hepatology uses Elsevier's.

Gut publishes significant liver disease content alongside GI research. Its BMJ Publishing Group policy is consistent with what Hepatology requires. If you've submitted to Gut, the AI disclosure expectations will feel familiar.

Gastroenterology also publishes hepatology research, particularly MASH/NAFLD studies, hepatitis trials, and liver cancer screening studies. Its AGA/Elsevier framework produces equivalent author requirements to Hepatology's AASLD/WK framework.

Liver International represents the Asia-Pacific perspective through APASL. Wiley's AI policy provides the publisher framework. The core rules align with Western journals, though the emphasis on particular clinical contexts may differ given the different disease epidemiology in the Asia-Pacific region (higher hepatitis B prevalence, different liver cancer demographics).

All five journals agree on fundamentals. No AI authorship, mandatory disclosure in Methods, no AI-generated images, full author responsibility. The liver disease field has reached the same consensus that cardiology, oncology, and other specialties have: the basic rules aren't controversial anymore.

Plan your AI use before you start

I've said it at the top and I'll say it again: decide which AI tools you'll use before you begin drafting. This doesn't mean you can't change your mind, but having a plan means you can track everything systematically.

The fibrosis staging problem

Hepatology papers frequently involve fibrosis staging, whether through liver biopsy (Metavir, NASH CRN), elastography, or serum markers. If you're using AI to edit sections that describe fibrosis results, be extra careful. Large language models sometimes introduce subtle inaccuracies in staging descriptions, such as conflating F2 and F3 criteria or mischaracterizing the meaning of particular elastography cutoffs. Have a hepatologist specifically review any AI-edited text about fibrosis staging.

Coordinate international author teams

Hepatology papers often involve collaborations spanning multiple continents. A multicenter MASH trial might include investigators from the US, Europe, Asia, and Latin America. Language barriers make AI tools particularly useful for non-English-speaking co-authors, and that's exactly what the policy permits. But coordinate the disclosure. The corresponding author should ask every co-author about their AI tool use before submission.

Supplementary materials count

Many Hepatology papers include extensive supplementary tables, figures, and methods. The AI disclosure covers supplementary content too. If AI helped edit your supplementary methods or generate supplementary figure legends, include it in the disclosure.

Statistical code disclosure

If GitHub Copilot or ChatGPT helped you write R or SAS code for survival analysis, propensity score matching, or any other statistical procedure, that's AI-assisted code generation. Disclose it. This is particularly relevant for Hepatology papers that use complex modeling approaches, competing risk analyses for transplant outcomes, time-varying covariate models for disease progression, and multi-state models for liver disease natural history.