Artificial Intelligence Tools in Medical Writing – Boon or Bane?
Year: 2025
Type: article
Abstract: Artificial intelligence (AI) tools have revolutionized the entire landscape of medical research and academic writing. ChatGPT is one of the most familiar and widely recognized AI tools among the many that exist. ChatGPT and generative AI tools have unprecedented efficiency and ease and unimaginable potential in the field of medical research and publishing. Yet, they can be deceptive and cause havoc. As we learn to navigate this evolving landscape, authors, reviewers, and editors must remain vigilant and ensure the ethical use of these tools in medical manuscript writing. The International Committee of Medical Journal Editors, JAMA Network Journals, and the World Association of Medical Editors have all issued specific guidelines on the use of generative AI. Nevertheless, the extent to which AI can be used remains ambiguous. Strengths and Opportunities AI tools boomed in 2022, after large language models (LLMs), like ChatGPT and Claude, made interaction with AI easy and accessible. AI tools can help both authors and reviewers of medical literature. They can aid in idea generation, searching for, and identifying gaps in literature. They can help develop methodologies to answer research questions, recommend appropriate tests, analyze results, and navigate complex statistical analyses. AI can help in generating visual and tabular content, drafting captions and descriptions, or formatting. Improving diction, correcting grammar, rephrasing, summarizing, and improving clarity—especially for non-native English users—are some of the core functions of AI. More importantly, formatting references, a dreaded task for any author, has become simple and easy using AI. AI can thus prepare an entire research manuscript. For reviewers, AI can help in detecting methodological inconsistencies, detect potential plagiarism, and provide structured feedback. Recently, otto-SR, an AI system powered by GPT-4.1 and o3-mini-high, was used to reproduce and update 12 Cochrane systematic reviews in 2 days. At the rapid pace at which technology is evolving, medical writing looks destined to witness an amazing spectrum of opportunities soon. Weakness and Threats A recent study examined the scientific community’s understanding of, and perspective on the role of AI, in research and authorship. It surveyed corresponding authors who published in the 15 highest-impact medical journals in 2022. Surprisingly, only one-third of authors were confident that AI would have a lasting presence in the field. More than two-thirds expressed concerns about content inaccuracies and the ethical implications associated with the use of AI. People also struggle with the “black box effect” in AI as the process by which AI models arrive at their decisions is opaque and unclear. Studies have highlighted several challenges with content created by LLM-based generative AI tools. Hallucination is a phenomenon where AI offers convincing, but completely fabricated, nonsensical, or factually inaccurate content in response to user prompts. AI tools are susceptible to the Dunning–Kruger effect, where people (or software) with limited competence frequently overestimate their competence. AI tools can perpetuate ‘citogenesis’, whereby an unverified or incorrect statement gains acceptance as fact through repeated citations. LLM-powered generative AI tools are notoriously poor at referencing medical literature. AI tools are trained on large datasets, and the quality of the AI’s results depends on the quality of data fed to it. If AI is trained on biased datasets, they can produce skewed or discriminatory results, leading to flawed conclusions, which can potentially undermine public trust in science. We also face a huge risk of deskilling the next generation of researchers. The ease with which manuscripts can be generated can encourage academic dishonesty and plagiarism. Reviewers also risk violating the authors’ confidentiality and proprietary rights by sharing manuscripts on AI platforms. Loss of confidentiality and legal copyright issues are other potential concerns. These issues only constitute the tip of the iceberg. What Lies Ahead? The future is both exciting and daunting, and we, in the medical community will have to navigate this uncharted territory with caution. Unbridled, we run the risk of being enslaved by our own creation. Using AI for idea generation and conceptualization is akin to forfeiting our inherent intelligence and questions the basic tenets of research. Research questions need to be our own, driven by our personal learning and experience. Any data generated by AI should be thoroughly vetted by human intelligence. While AI tools can be used for language refinement, generating images using AI should be avoided. Authorship comes with responsibility and oversight for the content, which AI cannot provide. AI tools cannot be, and should not be, listed as authors. Manuscript review involves critical assessment, which is often guided by years of experience. No reviewer should outsource their responsibility to AI. There is currently no technology that can reliably and accurately identify the use of generative AI in articles. Digital watermarking of AI-generated content appears to be an option but has its limitations. The best solution is personal integrity, which authors ought to display by honestly disclosing the nature and extent of AI usage, ideally in the article’s methodology section. AI tools carry great promise but have inherent risks. If used carelessly or deceptively, they risk undermining trust in scientific communication. The solution is not to ban them but to use them judiciously, with transparency, oversight, and accountability. Mankind is in pursuit of the Midas touch using AI tools. Like Midas, who gained the ability to turn all he touched into gold, AI has invaded every phase of our lives and made life easy. But Midas soon cursed his boon, when his food hardened into golden ice, and he starved. We must ensure that we do not let AI control us as humans may yet rue the day we created it. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.
Source: Neurology India
Author: R Girish Menon
Language: English
Subfield: Health Informatics
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W4414163971