European Heart Journal's AI Policy: ESC Rules for Cardiology's Top Journal

OUP publishes hundreds of journals. Its general AI policy requires disclosure in a cover letter and in the Methods or Acknowledgments. The ESC's own society-level guidance for its journal family adds clinical-data restrictions relevant to cardiology.

When OUP updates its guidelines or the ESC revises its position, EHJ reconciles them. Check the author instructions at the time of submission, don't rely on last year's version.

EHJ publishes substantial AI research, echocardiographic analysis, cardiac MRI segmentation, CT calcium scoring algorithms. If your paper is about an AI tool for cardiac imaging, that AI is your research subject, described in standard Methods.

The manuscript preparation AI policy is separate. If you developed a deep learning model for left ventricular ejection fraction estimation and also used Claude to polish your Discussion, these require two distinct disclosures:

• Research AI: standard Methods (model architecture, training data, validation)
• Writing AI: the AI disclosure statement

Don't combine them. Reviewers need to distinguish AI as a research tool from AI as a writing tool.

For clinical trial reports (anticoagulation strategies, heart failure drugs, device therapies) AI involvement in interpreting endpoints or drawing clinical conclusions isn't acceptable. You can use AI to improve prose. You shouldn't use it to draft sections where clinical claims are made. "AI helped me express my findings more clearly" is permitted; "AI helped me figure out what my findings mean" is not.

ESC Clinical Practice Guidelines carry the highest sensitivity. These documents determine how millions of patients across Europe are treated. The formal policy doesn't separate guidelines from original research, but editorial scrutiny is much higher. Undisclosed AI in a guideline paper would damage credibility far beyond the individual authors.

A proper EHJ disclosure includes three elements: the tool and version, what was and wasn't done with AI, and a statement of author responsibility. The OUP policy specifically asks authors to describe "the steps taken to validate the accuracy of outputs", so don't just name the tool and move on.

Original research example:

"During the preparation of this manuscript, the authors used ChatGPT (GPT-4, OpenAI) to improve the clarity of the Introduction and Discussion sections. All suggestions were reviewed by the corresponding author (A.B.) and senior biostatistician (C.D.). No AI tools were used in study design, data collection, statistical analysis, or interpretation of results. The authors take full responsibility for the content."

Imaging/AI research example:

"The deep learning model (EchoNet-LV) was developed using PyTorch and trained on the institutional echocardiographic dataset described in Methods. Separately, Claude (Claude 3.5, Anthropic) was used to edit the Results and Discussion for language clarity. These are entirely separate systems. All AI-suggested edits were reviewed by all authors."

Meta-analysis example:

"The authors used ChatGPT (GPT-4o, OpenAI) to improve readability of the Results section and check consistency of numerical values between tables and text. All outputs were verified against the original data extraction spreadsheets by two reviewers (E.F. and G.H.). The systematic search strategy, study selection, data extraction, quality assessment, and statistical analyses were performed entirely by the authors without AI assistance."

Notice what all three examples share: they name the specific tool and version, they draw a clear line between what AI did and didn't do, and they confirm author accountability. A vague "AI was used during manuscript preparation" won't satisfy EHJ's editors.

EHJ follows COPE guidelines. The escalation path:

• During peer review: The editor contacts you for clarification. Your paper gets flagged for closer scrutiny and possible AI detection screening.
• After acceptance, before publication: The paper can be pulled. You must provide a disclosure and explain the omission. This creates a record in Editorial Manager.
• After publication: Options range from a published correction to an expression of concern to retraction, depending on how extensively AI affected the scientific content.
• Institutional notification: In serious cases, the journal notifies your institution, potentially triggering a formal research integrity investigation.

Most omissions aren't malicious, authors didn't realize the policy applied to their use case, or they forgot to circle back and add the disclosure after using an AI tool midway through revisions. But "I didn't know" isn't a defense editorial boards accept in 2026. The policy is printed in the General Instructions, and checking those instructions is part of the submission process.

A 2025 study of 213 cardiovascular journals found that 170 (80%) had some AI policy, but only 3 of 17 high-impact journals actually mandated specific disclosure. None required standardized disclosure language. Here's how the top five compare based on their current author instructions:

Notable differences: EHJ's cover-letter-plus-body requirement is more explicit than most competitors. Circulation routes disclosure to Acknowledgments rather than Methods. JACC has a general AI policy but, as of the most recent study, doesn't mandate a specific disclosure section, though this may change as ACC updates its guidelines.

All five prohibit AI authorship and AI-generated images. Those two rules are universal across the cardiology journal landscape.

EHJ's readership is global, and many authors aren't native English speakers. The ESC represents cardiologists across dozens of countries where English isn't the primary language. The AI policy is explicitly designed to allow language improvement, that's the core permitted use case.

If English isn't your first language, you should feel comfortable using AI tools for grammar, sentence structure, and readability. That's what the policy permits. Just disclose it honestly. There's no stigma, it's the same function that professional language editing services have provided for decades.

The line to respect: don't use AI to generate scientific arguments, interpret data, or draft clinical conclusions. "Improving how I express my findings" is permitted. "Generating findings for me to express" is not.

Comply proactively if:

• You used any AI tool (ChatGPT, Grammarly, Copilot) during manuscript preparation
• The journal requires AI use disclosure in the methods or acknowledgments
• Your institution has its own AI use policy that may be stricter

Less concerned if:

• You used AI only for grammar/spell checking (most journals exempt this)
• Your use was limited to literature search or reference management