A Study of 635 Marketing Workers Ties Hiding AI Use to Fear, Not Bad Policy

David Torosyan spends part of his week reading other people's work and guessing which parts came from a chatbot. As the HR and payroll head at J&Y Law Firm in Los Angeles, he has watched staff turn around memos and case notes faster over the past year. What he hasn't watched much of is anyone saying out loud how they did it. "People are using it to write faster, organize smarter and communicate clearer," he told Digiday, "but quietly, often in the shadows."

That quiet workaround now has a number attached to it. A study published in June 2026 in the peer-reviewed journal Behavioral Sciences surveyed 635 marketing employees in Istanbul twice, six weeks apart, and found that hiding AI use after the fact is common enough to function as its own behavior, not just an occasional cover story. Researchers Filiz Mizrak, who studies management information systems at Atlas University in Istanbul, and Turhan Karakaya, of the business school at Dogus University, call it AI disclosure silence: using a tool like ChatGPT, Midjourney, or Canva AI to help produce a piece of work, then not telling anyone it was involved.

Fear beat confusion

In the Istanbul data, fear of negative evaluation correlated with disclosure silence at 0.56, and the sense that AI might make someone look less creative correlated at 0.58, the two strongest ties in the dataset. AI anxiety and worry about job security followed close behind. Trust in management and a clear written AI policy both pushed in the other direction, but more weakly, landing around negative 0.45.

Mizrak and Karakaya ran the numbers two ways: standard statistical modeling, then machine-learning models, Random Forest and XGBoost, interpreted with a method called SHAP that ranks which inputs actually moved the prediction. Both approaches pointed to the same four factors as the strongest predictors of concealment: fear of judgment, AI anxiety, perceived creativity threat, and how much an employee trusted management.

No single culprit, and a narrow sample

A separate test looked for one condition that had to be present for concealment to happen, the kind of single bottleneck a company could fix with one policy memo. It didn't find one. High disclosure silence showed up through several different combinations of fear, distrust, and unclear rules, which the authors say means there's no single lever to pull. The sample was narrow by design, too: all 635 respondents worked in marketing in Istanbul, split across ad agencies, e-commerce companies, and corporate marketing departments, so the exact numbers may not transfer cleanly to other industries or countries.

“Employees aren't hiding their AI use because they're trying to get away with something. They're hiding it because they're trying to get ahead without getting in trouble.”

Sam DeMase, a career expert at the job site ZipRecruiter, puts the fix on employers rather than employees. "Leadership has a responsibility to craft policies around practical and ethical AI usage on the job and communicate those policies with their team," she said, adding that transparency removes the fear and shame some employees attach to using AI. Justin Hale, a master trainer at the workplace-training firm Crucial Learning, makes a related point aimed at managers: go into the conversation expecting people to feel unsure or uneasy, and treat that as a reason to be more open, not less.

What this changes, if anything

None of this means AI use at work is rare. The same Cornerstone OnDemand survey put adoption at 80 percent of U.S. and U.K. workers, while only 44 percent of U.S. employees said they'd received any training on how to use it. That combination, high use paired with low training, closely matches the setup the Istanbul study flagged as predicting silence: people using a tool they were never taught to use well, inside workplaces that haven't said clearly whether that's a problem.

Employers, for what it's worth, mostly don't see themselves as the obstacle. In the same Cornerstone survey, 64 percent of U.S. employees and 62 percent of U.K. employees said their organization broadly supports AI use, even though only a third reported being often or always encouraged to use it in practice. That gap between a stated open-door policy and what employees actually feel day to day is where disclosure silence tends to live, according to both studies.

The overlap between the two studies points to policy clarity and trust, the two factors that moved disclosure in the opposite direction from fear in the Istanbul data. Neither one erases the underlying anxiety by itself. But a workplace that never states out loud whether using AI is allowed is, by this research, the one most likely to end up with people using it anyway and simply not mentioning it.