We spend a lot of time telling women to lean in to AI. To upskill. To get over their hesitation. The gender AI gap is framed, almost universally, as a problem women need to fix in themselves.

The gap is well documented. The why is not.

Research has consistently shown women adopting AI at rates 20% lower than men, and this gap persists even when skills are equivalent. While adoption rates will continue to evolve, the barriers remain consistent. Harvard Business Review identified the top three as trust, ethics, and fear of judgment. Trust and ethics are precisely the qualities AI needs most - women's hesitation is not a weakness, it is a signal. But it was that third barrier, fear of judgment, that I wanted to test. If women believe they will be judged more harshly for using AI than their male counterparts, are they right to be afraid?

So I decided to find out.

At Code For Good Now, we conducted a nationally representative study of 1,000 UK adults. Each person was shown the same AI-supported CV for a marketing role. Respondents were told the candidate had used AI. One thing only changed: the name at the top. Half saw Emily Clarke. Half saw James Clarke. We measured candidate strength, trust, competence, and sentiment. The results were unambiguous.

Same CV. Same AI use. Different judgment.

Reviewers of the CV were 22% more likely to question whether Emily could be trusted. They were twice as likely to doubt her competence. The fear women have about being judged for using AI is not perception. It is measurable. It is real. And we are calling it the AI Judgment Penalty.

The difference shows up most starkly in the language people actually used. Reviewing James' CV, respondents said he "just needed a bit of help putting it together." Reviewing the identical CV for Emily, the response was different: "She can't even write a CV herself - not sure she has the skill to carry out the job."

Same CV. Same AI use. When a man uses AI, we question his effort. When a woman uses AI, we question her integrity.

The penalty was felt across all demographics. Notably, male Gen Z respondents, the generation that grew up with AI and uses it most, were 3.5 times more likely to describe Emily's CV as weak.

This matters for the comms industry in particular. AI is now embedded in how briefs are written, how creative is evaluated, how campaigns are optimised and measured. The conversation about AI in this industry is urgent and ongoing. But there is a question being asked too quietly: who carries the cost of using it?

When a man uses AI, it reads as efficiency. When a woman uses the same tool in the same way, her trustworthiness is questioned and her competence is doubted. That asymmetry doesn't just affect how individuals feel about AI, it shapes whether they use it at all.

AI is shaped by its userbase. The input determines the output. Fewer women using these tools means fewer women's perspectives, experiences and needs reflected in what AI produces, and in how it evolves. We are not just talking about who benefits from AI. We are talking about what AI becomes.

AI literacy training won't close this gap on its own. You cannot upskill people out of a structural bias. If the culture around AI use is unequal, if women are penalised for the same behaviours that earn men credit, then investing in access and tools without addressing that dynamic is building on sand.

This is why we designed Permission to Prompt, our training programme built specifically around the barriers we are measuring. Not just how to use AI, but what stops people from using it in the first place. It addresses the judgment layer directly, helping organisations build consistent standards for AI use, create cultures where experimentation is protected, and ensure that AI is evaluated and adopted equitably across teams.

Who shapes AI shapes the future. The hesitation women feel around AI is not their problem to fix. It is a rational response to a real penalty. The research proves it. The question now is whether the industry is willing to act on it.