Same Draft. Different Verdict.

Three recent studies indicate that the consequences of using AI at work vary depending on the user (Harvard Business School AI Institute, 2026; Lean In, 2026; Fortune, 2026). For marketing leaders, a "low adoption" rate may actually reflect a team that has learned to withhold information (Spin Sucks, 2026).

Most marketing teams conducted similar surveys this year, asking about AI usage frequency and applications. When results were lower than expected, the typical response was to implement additional training, upgrade tools, or schedule informational sessions.

The research suggests the survey was measuring something other than usage.

The penalty for AI use is real and unevenly distributed.

Start with the cleanest experiment. In May, Zehra Chatoo, a former Meta strategist who now runs the think tank Code For Good Now, showed 1,000 UK adults an identical résumé with the same disclosure that AI had helped build it. The only variable was the name at the top: James Clarke or Emily Clarke.

James got a 97% approval rating. Emily's résumé was rated strong by 76%. Reviewers were 22% more likely to question whether Emily could be trusted, and twice as likely to doubt she could do the job. Gen Z men called her résumé "weak" 3.5 times as often as they called his.

The free-text comments reveal the underlying bias. James’s AI use was interpreted as a minor need for formatting assistance, while Emily’s was seen as evidence of incompetence. As Chatoo summarizes: when women use AI, "we question their integrity."

The same document and disclosure led to different outcomes.

The people being judged already know

That penalty is not a surprise to the people it lands on.

Lean In surveyed 1,015 US adults in March. Twenty-nine percent of women said they worry they'll be perceived as cheating when they use AI at work, against 22% of men. Among people who had used AI on the job, 23% of men had been praised for it, versus 18% of women. Men were also more likely to report that their manager encouraged them to use AI — 37% to 30%.

Taken together, these findings show that women anticipate a penalty, receive less recognition and encouragement, and consequently use AI less frequently: 33% of men reported daily or constant use, compared to 27% of women.

The Harvard Business School working paper that pulls the global picture together makes the same point at scale. Katelyn Cranney, Solène Delecourt, and Rembrand Koning reviewed 76 sources across more than 100 countries and 300,000 people, updated in August. Weighted adoption: 47.8% of men, 39.3% of women. The relative gap narrowed since 2023, then stalled at roughly 16% since early 2025.

The gap holds inside the same job. At one global technology company, 43% of male software engineers had used the internal AI coding tool at least once, versus 31% of female engineers. And the paper cites research on engineers where identical AI-assisted work drew lower competence ratings when the engineer was a woman — a penalty, the authors note, that the women had correctly anticipated.

The researchers identify five factors contributing to the gap. Knowledge and perceived usefulness have improved as AI tools become more widespread. However, institutional support, social legitimacy, and trust remain unchanged and are influenced by workplace culture.

What this does to your adoption number

Here is the part that belongs to marketing leaders specifically, and it isn't about hiring.

If a significant portion of your team anticipates judgment for using AI, they may underreport their usage on adoption surveys. They might describe daily use as "occasional" or omit the tool from process documentation. As a result, surveys indicate low adoption, prompting leadership to invest in training, yet the numbers remain unchanged because the underlying issue is not a lack of skill.

Gini Dietrich at Spin Sucks described exactly this pattern with a client of 11,000 employees: monthly self-reported AI usage remained low and predominantly male, while platform logs showed nearly universal daily use. As she notes, "Adoption isn't low. Admission is low".

Marketing and communications teams are most affected due to structural factors. These functions, which rely heavily on AI for drafting, research, and analysis, are predominantly staffed by women, while those developing AI policies are often not. Consequently, the group most likely to use AI daily faces the greatest penalty for disclosing it, while those least penalized determine what constitutes "real" work.

This dynamic leads to a lack of shared practices, as individuals are reluctant to discuss their methods. There are no established quality standards for AI-assisted work, since official reports understate its prevalence. Leaders may see survey results indicating 20% adoption, while license data shows 90%.

The measurement problem is a management problem

I've written before that most marketing AI problems turn out to be ownership problems, not tooling problems. This is the same finding from a different angle. The gap between self-reported use and logged use isn't noise. It measures how safe people feel describing their process. That number is worth tracking on its own.

Several recommendations follow.

Standardize disclosure practices. When only one person notes AI use on a draft, they become subject to scrutiny. If every draft includes a brief process note detailing the tool used, retained content, revisions, and checks, the focus shifts from the individual to the workflow, as intended.

Focus reviews on output quality. Chatoo's data indicates the penalty affects perceived competence. Therefore, assess whether the brief meets requirements and inquire about process consistently for all team members, or not at all.

Compare survey responses to platform usage data. Identify discrepancies and determine which individuals fall within this gap. This reveals the true adoption issue, which is rooted in trust.

Monitor follow-up interactions. Over the next month, observe who is questioned about authorship and who receives positive feedback. The praise gap identified by Lean In—23% for men versus 18% for women—can be addressed through increased awareness and intentional recognition.

The frame to hold

While it may be tempting to categorize this as a diversity issue and delegate it to HR, this is fundamentally a measurement challenge within your function, and the solution lies within your team.

A team that conceals its most valuable new skill isn't developing a capability; it's creating a liability. This becomes evident when questions arise about why those recognized as AI-fluent are consistently labeled as resourceful.

Less theater. More outcomes. Starting with an honest number.