Hi Reader,

What if I told you that I used AI to write this article for you today? Would you continue reading it? Probably not!

Or perhaps you would continue, but with a little more caution.

You might wonder whether the ideas are truly mine. Did I verify the evidence? Are the references real? Did I protect confidential information? And, perhaps most importantly, did I contribute enough thought and expertise to deserve your trust?

These questions capture much of the discomfort medical writers feel about AI. The fear is not simply that AI will “take our jobs.” It is also the fear that using it will make our work appear less thoughtful, less original, or less trustworthy. Many writers I talk to also describe a kind of AI stigma: This sense of embarrassment or hesitation around admitting AI use, as if it signals a lack of skill or effort. Even when AI is used appropriately as a support tool, there can be a worry that others will judge the work more harshly or question the writer’s competence, which can make transparency feel surprisingly personal and emotionally charged.

As usual, let's talk evidence!

AI use is growing, but transparency is lagging

In an analysis of 105,538 manuscripts submitted to 13 JAMA Network journals, only 3.3% disclosed using AI. However, reported use increased from 1.7% to almost 6% between 2023 and 2025. Among those who disclosed AI use, 67.7% used it for language correction or refinement, while only 5.5% reported using it for manuscript drafting.

Similarly, 5.7% of research submissions to 49 BMJ journals disclosed AI use. The authors noted that this was considerably lower than survey estimates, suggesting that AI use may be underreported because writers are uncertain about the rules or concerned about how disclosure will be perceived.

This means writers are using AI, but many are still uncomfortable admitting it.

A quick question for you

Why you should be cautious

The concerns around AI are real. We know so far that AI tools can produce confident but inaccurate statements, fabricate believable references, flatten an author’s voice, and expose confidential material when content is entered into an external system.

A large analysis of scientific production found that researchers adopting LLMs increased their publication output by 23.7% to 89.3%, depending on their field and background. But the analysis also found that more sophisticated language did not necessarily correspond to stronger scientific content. In other words, polished writing can create an impression of quality that the underlying work has not earned.

What has changed at JAMA?

Here is a useful resource every medical writer should read. In August 2026, JAMA Network published updated guidance explaining where AI use is acceptable, where it must be disclosed, and where it is not permitted.

If you want to keep this guidance handy, you can download our practical infographic summarizing this updated JAMA Network guidance on AI use in medical publishing.

Remember that this is JAMA Network guidance, not one universal policy for every publication. Always check the target journal’s current instructions. The ICMJE recommendations also emphasize human accountability, transparent disclosure, confidentiality, and the principle that AI cannot qualify as an author.

A resource you may enjoy

One of the main reasons writers turn to AI is simple: it can help us work faster and save time. When used thoughtfully, it can support tasks such as organizing ideas, refining language, and managing repetitive work, while leaving scientific judgment and final responsibility with the writer.

My takeaway

Using AI does not automatically make a writer careless, and refusing to use it does not automatically make the work more trustworthy. When used thoughtfully and responsibly, AI can also help writers save time and improve their productivity.

The real questions, in my opinion, are:

How was it used? Was it appropriate for the task? Was it disclosed when required? And did a qualified human remain responsible for every word, claim, and reference?

For us, medical writers, AI should support judgment, not replace it. Our greatest value will continue to come from the parts of the work that require scientific understanding, ethical responsibility, audience awareness, and human insight. At the same time, we need to acknowledge the emerging issue of shame around AI use where writers may feel pressure to conceal assistance rather than disclose it.

I still feel this pressure myself. Sometimes, I worry that we are being pushed toward greater AI use before we have fully considered what we might lose along the way. That uncertainty makes me hesitant to rely on it more heavily.

Will clear, normalized disclosure helps reduce this stigma and encourages responsible use? I'm not sure, but all I know is transparency is part of ethical scientific communication, not a sign of diminished authorship.

And yes, AI assisted me in developing this newsletter. I used it to help organize the discussion and review possible sources. I then evaluated the evidence, checked the references, refined the interpretation, created the infographic, and accepted responsibility for the final content.

Your turn

Do you ever feel hesitant or uncomfortable admitting that you use AI?

Have you ever felt ashamed or judged for admitting that you use AI?

Where do you currently draw the line when using it in your medical-writing work?

Reply to this email and tell me. I would genuinely love to hear how you are navigating it.