The Responsibility of Authorship in the AI Era

The debate about AI in scientific writing is becoming increasingly confused. Should authors disclose that they used AI to improve a manuscript? The Journal of the American College of Cardiology (JACC) has taken a clear position.¹ The journal maintains that AI use doesn’t need to be disclosed when AI is used solely as a writing tool.

I think they get the fundamental point right. AI can correct grammar, restructure a paragraph, improve clarity or help organize an argument. However, drawing a reproducible line between “minor editing” and “substantial rewriting” is becoming almost impossible. More importantly, disclosure answers the wrong question.

The real question is responsibility. What matters is not whether an AI model touched the prose. What matters is whether the authors understand the science, verify the claims and references, remain intellectually engaged, and take full responsibility for every word they submit.

The Reality About AI Disclosure

The Journal of the American College of Cardiology therefore makes an important distinction. The journal maintains that AI used for writing is a tool whereas AI used to conduct research is a method that must be reported.

That seems much more meaningful than another checkbox, especially because disclosure rules appear remarkably difficult to enforce. In one analysis, around 70 percent of more than 5,000 journals had AI policies yet only about one in 1,000 papers publicly disclosed AI writing assistance despite evidence suggesting substantially broader use.

What is My Take on AI in Scientific Publishing?

The problem with AI in scientific publishing is not AI writing. The problem is AI replacing intellectual work. A beautifully written discussion based on poor science remains poor science. A fabricated reference remains fabricated whether it came from ChatGPT, a careless researcher or an exhausted fellow. Authors cannot outsource responsibility to an algorithm.

There is also an important equity issue. AI writing tools can reduce the linguistic disadvantage faced by researchers who are not native English speakers. Requiring them to label that assistance while native speakers obtain the same advantage naturally deserves careful consideration.

So perhaps we should spend less time asking: “Did AI write this sentence?” and much more time asking: “Is the science sound? Are the references real? Are the conclusions justified? Does the author stand behind every claim?”

AI is a tool. Authorship is responsibility. That distinction matters far more than disclosure.

Dr. Cademartiri is the director of advanced cardiovascular imaging and photon-counting CT at the Scientific Institute for Research, Hospitalization, and Healthcare Synlab Diagnostic Network in Naples, Italy. He is also a consultant in advanced cardiovascular imaging at CDI/Centro Diagnostico Italiano in Milan, Italy.

Reference

1. Krumholz HM. AI in medical writing: why JACC does not require disclosure. J Am Coll Cardiol. 2026;88(14):1687-1689. Doi: 10.1016/j.jacc.2026.08.032.

(Editor’s note: This blog is adapted with permission from Dr. Cademartiri’s original LinkedIn post available at: https://www.linkedin.com/in/filippocademartiri/recent-activity/all/ )