Saying how AI helped

Many people assume that admitting AI helped can only cost them, so the safe course is to say nothing. Half of that is right, and it is worth knowing which half.

The cost is real. In 2025 two researchers at the University of Arizona’s business school, Oliver Schilke and Martin Reimann, reported 13 experiments in which people judged someone else’s work. The work varied: a professor grading, a job applicant’s letter, an advertisement, a performance review, a work email.¹ Across the experiments, people who disclosed using AI were trusted less than people who didn’t.¹ Careful wording didn’t remove the effect. Some notes said a person had reviewed and revised the work, or that AI was used only for proofreading, and trust still fell.¹ The drop was smaller, but not gone, among people who liked new technology or believed AI to be accurate.¹

The other half is wrong. In one of the experiments, people imagined filing their taxes online and chatted with a tax advisor who prepared an estimate of their refund. Some then read that the estimate had been prepared with generative AI. Others read a news item, based on an anonymous leak, revealing that he used it for all his estimates. A third group was told nothing. Trust was highest when nothing was said, lower when he disclosed it, and lowest when someone else exposed it.¹ In other words, silence did best only when nobody found out. These were controlled experiments, not studies of real workplaces, but the pattern held across them.

And in a great deal of writing, silence isn’t a choice you get to make. Schools, journals and many employers now have rules, and those rules ask you to say how AI helped.

What the rule books say

The major style guides each publish advice. These are dated rules: what follows is as of late 2026. The Chicago Manual of Style says you must credit an AI tool whenever you use text it generates. For most writing, naming the tool in your text is enough; for formal work, such as a student paper or a research article, it suggests a numbered note.² If you edited the AI’s text, Chicago says to say so.² APA says that if you used AI when writing a paper, “it will usually be necessary to disclose the use and cite the chat and/or AI tool.”³ MLA says not to treat the AI tool as an author.⁴

Groups that set rules for scholarly publishing agree on the central point. AI tools can’t be authors, because they “cannot take responsibility for the submitted work,” COPE says, and authors must disclose which tool they used and how.⁵ The International Committee of Medical Journal Editors says the same: chatbots should not be listed as authors, and their use must be disclosed.⁶

Spot it today: these pages change. In 2023, APA advised citing only the AI tool, because a particular chat couldn’t be retrieved later. By 2025, with chats that can be shared by link, it showed how to cite the chat itself.³ APA also notes that what is allowed, and how to disclose it, depends on instructor, department and publisher policies.³ Before you write the note, check the current guidance of the place your work is going: your school, your employer, your publisher.

People often ask whether they could simply be caught by a detector. The Hustler ran the Vanderbilt email through a tool OpenAI had released to spot AI writing, and the tool rated it likely AI-generated.⁷ By OpenAI’s own figures, in its tests the tool labeled 26 percent of AI-written text “likely AI-written” and wrongly flagged human writing 9 percent of the time. In July 2023 OpenAI withdrew it “due to its low rate of accuracy.”⁸ Chapter 15 comes back to detectors.