2026 Lancet AI Guidelines: Do’s and Don’ts Before You Submit

Have you noticed these declarations at the end of recently published papers?

AI Use Disclosure

ChatGPT was used for reference formatting purposes only.

Declaration of Generative AI Use

ChatGPT 5.5 [Open AI] was used solely for language refinement, idea exploration, and proofreading. The authors take full responsibility for the content and originality of the manuscript.

Declaration of Generative AI and AI-Assisted Technologies in the Writing Process

While preparing this paper, the authors used Grammarly and ChatGPT solely to improve the grammar and correct the English of the manuscript. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the final content of the publication.

This is because journals, and even conferences, increasingly require authors to include a statement on artificial intelligence (AI) or large language models usage as part of their anonymous submissions.

Assistive AI vs. Generative AI

Did you know that AI and AI-assisted tools have been a part of research workflows and academic publishing long before the current AI boom? It’s true! Machine learning-based software such as reference management systems, statistical packages, and grammar-checking tools has long been an integral part of a researcher’s toolkit. In fact, universities and journals rely on algorithm-based plagiarism-detection tools to help maintain academic and research integrity. Then why are authors now being asked to disclose their use of AI and the integration of AI-generated outputs into their writing process?

It’s because AI is no longer limited to assistive technology. It has evolved to acquire a generative capacity, and generative AI can introduce bias and produce inaccurate fabricated content, compromising validity and reliability. The launch of ChatGPT version 3.5, a conversational AI chatbot, for use and exploration by the public in November 2022 marked the boom of AI into the mainstream. Since then, generative AI tools have advanced leaps and bounds to not just produce text but also generate images and datasets. While AI can be quite useful, it is not perfect. AI models often hallucinate, i.e., they can fabricate completely coherent academic manuscripts, replete with datasets, patient information, convincing conclusions, and entirely fake results and references. This poses a serious threat to research and medical integrity, and without ethical practices in place, misinformation and fabricated findings could quietly work their way into the literature.

Do’s and Don’ts: Responsible AI Use in Research Writing

The internationally acclaimed Lancet Group, comprising leading medical, global health, and clinical health journals published by Elsevier, directly addresses AI adoption with a dedicated, in-depth granular section titled Use of generative AI and AI-assisted technologies in the latest 2026 Lancet Guidelines for Authors. These AI-specific guidelines emphasize the renewed responsibilities of authors when it comes to using a new and rapidly evolving technology during both the research process and manuscript preparation.

Here’s a quick overview before we get into the details:

First, let’s go over what authors should do:

Include an AI declaration statement: Upon submission, authors should include a separate AI-specific declaration statement disclosing the generative AI or AI-assisted tools used while preparing the manuscript. This statement should be provided at the end of your manuscript, and it will appear as such when published.  The use of AI for basic grammar checks and proofreading doesn’t require a declaration.

This is the baseline requirement and several researchers who have sounded the alarm on how AI is being adopted in research work faster than policies can keep up¹. As AI use continues to gain prominence in scientific research, detailed disclosures regarding how and why AI and AI-assisted tools were used in research workflows and manuscript writing processes are going to be essential for establishing transparency and maintaining research integrity.

Document your use of AI: Maintain a detailed record of the AI tool used. Lancet specifically asks authors to document three pieces of information: [i] the name and version of the AI tool (s) used, [ii] the purpose for which AI was used (e.g. formatting or restructuring sentences for readability), and [iii] the extent to which authors reviewed the AI-generated content.

The first two are easy asks. It is easy to record these details in your research notes, and it should be standard practice. You can refer to it if questioned during reviews or post publication. The last bit, however, is particularly important as it requires authors to engage with AI tools not just as users but as experts who will review, edit, and verify the output produced by AI. For example, did you check the AI text for factual errors or bias? Authors must retain complete responsibility for the accuracy of the information included in their submission.

Provide prompts where applicable: Authors must not only keep a record of the prompts used to obtain the desired AI-generated output but also provide them in the manuscript where applicable. The prompts used can also be requested by editors as they see fit.

Disclose AI use in the Methods section: Describe in detail how you used AI in your research process (not the writing process; AI use in the writing process should be declared at the end of the manuscript) in the Methods section.

This can include AI-enabled tools within legacy software such as data analysis packages. Other examples include details of AI platforms used to directly translate text to or from another language or of AI tools to create experimental stimuli. 

Review the terms and conditions of AI tools: Authorsshouldcarefully read theterms of service of the AI technology being used. Lancet explicitly asks authors to ensure that the terms and conditions grant AI the right to use the author-entered content solely for providing the intended service. The AI tool should not have any other rights on the author’s content, including automatic inclusion of the entered material in further training the AI. The AI platform should maintain the privacy and confidentiality of the data entered. Authors must also ensure that the AI tool of choice does not come with restrictions on how the AI-generated output can be used as this can create issues during journal publication.

This guideline is an important prerequisite for using AI in academic writing. Authors must be aware of the various clauses included in the terms and conditions. AI tools should not gain ownership over the data entered or the output generated as this can clash with journal publication polices.

Moving on to what should be avoided:

Don’t use AI as an intellectual substitute: The use of AI tools in manuscript preparation must be solely assistive. Authors should not employ AI as a substitute for human reasoning skills and expertise.

AI is not perfect and many AI platforms come with the disclaimer that AI can make mistakes and that all responses must be double-checked.

Don’t list AI as a co-author: Never list an AI tool as an author, co-author nor cite AI as an author. 

This is because authorship includes taking responsibility for the tasks that went into preparing the manuscript and no technology can assume that responsibility. The responsibility of conducting research and preparing a manuscript rests with humans alone.

Don’t use AI to directly generate figures: AI should not be used to directly create any visual components of the manuscript. The only exception is when articles focus on demonstrating or testing AI capabilities in generating artwork (graphs, models, figures, etc.) as part of the research question.

This guideline is in place to avoid potential copyright issues arising from using unlicensed or uncredited image sources on the internet. However, the use of generative AI tools for exploring ideas and concepts for image creation faces no such restriction.

Don’t generate duplicate or copyrighted images:Authors are required to ensure that their use of AI does not include generating replicas of existing copyrighted visuals, real individuals, or identifiable brands. Nor should there be any resemblance to an individual’s voice.

In addition to these dos and don’ts, authors must accept that journal editors have the final say on determining whether the use of AI is appropriate and that they may, at any time, request further details about AI use.

Author Accountability in AI Use

The Lancet AI guidelines explicitly require authors to practice both responsibility and accountability. You can’t have one without the other.

Author accountability means taking ownership of the information included in the manuscript and being answerable for it. In the context of AI use, it means:

• Thoroughly reviewing AI-generated outputs for accuracy, coherence, and impartiality
• Editing and incorporating AI-generated material to represent the author’s original voice and contribution
• Practicing complete transparency about using any tools or sources, AI or otherwise.
• Ensuring that the submission-ready manuscript does not violate any data privacy, intellectual property, or other rights.

The underlying premise of these guidelines is that authors must adopt a mindset that emphasizes documenting and disclosing AI use in order to maintain and promote transparency and trust between authors and editors, peer reviewers, and readers while complying with AI’s terms of use.

Key takeaways:

• AI can only assist, not replace human critical thinking
• Authors are responsible for describing, documenting, and disclosing AI use.
• Authors are accountable for ensuring the accuracy and impartiality of AI-generated content.
• Journal editors have the final say in determining whether the use of AI is appropriate.

Additional Reading:

1. Resnik, D. B., & Hosseini, M. (2025). The ethics of using artificial intelligence in scientific research: new guidance needed for a new tool. AI and Ethics, 5(2), 1499-1521.