Cautions for the medical writer using AI tools

As much as artificial intelligence (AI) tools have transformed medical writing, they have brought with them a host of cautions that writers must take into account. Using AI tools responsibly helps writers maintain the quality and originality of their work and prevent future conflicts.

Possibility for bias and inaccuracy

Popular generative AI (genAI) tools, such as ChatGPT, Gemini or Claude, rely on large language models (LLMs) to produce answers. Thus, an AI tool or chatbot is as impartial as the training data that define its algorithm. Many research papers also sit behind expensive paywalls, meaning AI tools cannot access them. These factors lead to non-inclusive and biased answers.

AI tools can confidently present incorrect information about medical procedures, drug dosages or treatment protocols as correct, without giving any indication of error. They might mix up similar-sounding medical terms or make subtle mistakes that only subject matter experts would notice. In addition, they may also ‘hallucinate’ or present invented claims as hard facts.

What you can do as a medical writer

Fact-check everything. Always verify AI-generated content against multiple trusted clinical guidelines and medical databases, such as PubMed or Google Scholar.
Keep a list of subscription-based journals. Manually check for any research that the AI might have missed due to paywall restrictions.
Create topic-specific checklists. Checklists help to identify AI errors in your field (eg, related to drug names, dosages or procedure steps).

Confusion about authorship and citation

The unclear role of AI-generated content in publications opens up a number of dilemmas. Should genAI be listed as an author? Many publishing guidelines do not address AI contributions. There is also the dilemma of who or what to cite; AI tools often blend information from multiple sources, thus creating confusion with referencing.

Medical writers might struggle to verify original sources or provide proper citations for AI-generated content. The situation becomes especially problematic when writing for peer-reviewed publications or creating content that will later undergo thorough reference checking.

What you can do as a medical writer

Document your AI tool use. Keep records of when and how you used AI while writing each piece.
Create a clear trail of your sources. Save links and references as you go, rather than tracking them down later.
Consult journal and/or client requirements. Different journals or clients may have different recommendations regarding AI use and declaration.

Ethical concerns

When using AI tools, medical writers must navigate strict regulations regarding patient data protection and copyright laws. AI tools can compromise patient privacy with improper handling. When these systems process large amounts of sensitive medical data, small mistakes could lead to privacy breaches and break data protection laws.

The legal landscape around AI and plagiarism is equally tricky. Existing guidelines do not precisely define if AI-written medical content counts as original work or if it breaches someone else’s copyright. This legal grey area causes headaches for everyone involved in medical writing.

What you can do as a medical writer

Use anonymised examples and case studies. Remove potential patient identifiers when giving information to generative AI tools.
Check for AI guideline updates. As a writer, you should stay updated with the latest developments in AI-related regulations.
Create a personal ‘copyright-clear’ database. This database can include resources that you can safely use with AI tools without copyright concerns occurring.

Moving forward

As AI tools become a bigger part of content creation in healthcare, they will require strict legal frameworks to regulate their use. While the current AI tools offer benefits for medical writers, their limitations and risks need to be carefully considered. Medical writers should see AI only as a facilitator of their work. They are responsible for understanding the shortcomings of AI tools and addressing them promptly.

Cautions for the medical writer using AI tools