A UC Berkeley professor’s acknowledgment that AI helped edit a published op-ed is highlighting an increasingly difficult question for journalism, academia and professional writing: where does ordinary AI assistance end and authorship begin?

WHAT’S HAPPENING

UC Berkeley mathematics professor Zvezdelina Stankova recently published an op-ed in the San Francisco Standard arguing that some incoming students are arriving at the university without the mathematics preparation traditionally expected for higher-level coursework.

After journalists from Berkeley’s student newspaper questioned whether AI had been involved in producing the article, Stankova acknowledged using artificial intelligence as part of the editing process.

She said the research, analysis and overwhelming majority of the work were performed by humans, while AI was used to assist with editing and gathering relevant material.

The San Francisco Standard subsequently reiterated that AI may be used as a tool, but human authors remain responsible for the accuracy, reasoning and content of anything published under their names.

The controversy has created a second debate largely separate from Stankova’s arguments about mathematics education:

How much AI assistance can a writer use before readers should consider AI part of the authorship process?

WHY IT MATTERS

AI-assisted writing increasingly exists on a spectrum.

Most people would not consider spell-check to be co-authorship.

Grammar suggestions are generally treated similarly.

But the boundaries become less clear when AI begins:

rewriting sentences,

restructuring arguments,

summarizing research,

gathering supporting material,

or generating substantial portions of prose.

That creates an emerging disclosure problem.

A writer may genuinely develop the argument, conduct the analysis and remain completely responsible for the finished work while still using AI extensively during production.

At the same time, readers may reasonably want to know when technology played a meaningful role in shaping something presented as an individual’s writing.

The central issue may therefore become less about whether AI was used at all and more about what role it played.

WHO BENEFITS

Writers and researchers can use AI to improve clarity, organize information and accelerate portions of the editing process.

Publishers may benefit from faster workflows when clear standards preserve human accountability.

Readers could benefit from greater transparency if publications establish understandable rules around meaningful AI assistance.

AI companies benefit as writing and research tools become integrated into everyday professional workflows.

WHO LOSES

Writers who rely heavily on AI without disclosure may face questions about credibility if the extent of that assistance later becomes public.

Publishers without clear AI policies may be forced to make judgment calls after controversies emerge rather than establishing expectations beforehand.

Readers could lose confidence if they cannot determine whether a published argument primarily reflects the named author’s reasoning or machine-generated material.

There is also a risk in the opposite direction.

Treating every use of AI as equivalent to AI authorship could unfairly stigmatize routine editing assistance that does not meaningfully replace human thought.

WHAT HAPPENS NEXT

Expect publishers, universities and professional organizations to move away from simple rules asking whether AI was used.

The more important questions will likely become:

What did the AI actually do?

Did it generate the argument or merely improve the wording?

Did a human verify the research?

Who is accountable for errors?

At what level of assistance should disclosure become necessary?

AI detectors are unlikely to completely resolve the issue because they can produce uncertain or disputed results.

That makes human accountability and transparent editorial standards increasingly important.

The larger shift is already underway.

AI is becoming ordinary enough that simply asking:

“Did you use AI?”

may eventually become the wrong question.

The better one may be:

“How much of the thinking was yours?”