AI Jobs Pay More, and Women Are Getting Fewer of Them

What the new numbers say

The Guardian's investigation by Arielle Pardes, published on October 4, 2026, puts a single figure at the centre: across AI occupations, men's median pay is $45,000 higher than women's. Part of that comes from the types of AI jobs men are more likely to hold. The same reporting notes that women are more likely to work in roles highly exposed to AI disruption, such as customer service.

Much of the hiring picture comes from LinkedIn. Women made up only 26% of US hires into jobs requiring AI skills, against 50% of hires into non-AI roles. These are also the jobs where the money is moving: AI postings on LinkedIn have roughly doubled since 2023, and the typical listing offers about $177,000, compared with about $80,000 for a non-AI role.

Put those together and you get a compounding problem. The fastest-growing category of work pays more than twice as much, and women are entering it at about half the rate they enter everything else. Even a company that pays fairly within every job title can end up with a widening pay gap simply because of who lands in which titles.

The gap is about which AI jobs, not only whether

It would be easy to read the headline as "fewer women in AI" and stop there. The more useful detail is how the split looks inside AI work. According to Axios, 82% of 2025 hires into "member of technical staff" AI roles were men, with a median listed pay of $223,000. Data annotator jobs, at a median of $51,000, were split roughly 50/50.

The top of the ladder looks worse. Women accounted for 20% of Head of AI hires, 26% of Director of AI hires and 18% of Member of Technical Staff hires. Typical pay for those roles runs to about $236,000, $300,000 and $245,000 a year respectively. Across 27 countries, women hold only 13% of leadership roles at AI firms, compared with 20% of the same roles in the wider tech industry.

LinkedIn frames this as three stacked barriers: women's share of AI roles is about 10 points lower than in non-AI roles, their share of the workforce is about 5 points lower at AI-focused companies, and the gender gap is about 15 points wider at C-suite level. For an employer, the lesson is that headcount ratios hide the problem. You need to look at which AI work women are doing, and at what level.

Giving everyone a license is not a policy

Most companies will not hire a Head of AI this year. What almost all of them are doing is rolling out AI tools to existing staff, and this is where the research should worry managers. A Harvard Business School review of 76 sources covering more than 300,000 people found a weighted AI adoption rate of 47.8% for men and 39.3% for women. Women also spent less time per visit on six of the top ten AI tools in a US panel.

The finding that matters most for employers is that the gap survives inside the same workplace. A PNAS survey of 18,000 Danish workers in AI-exposed occupations found women were 16 percentage points less likely to have used ChatGPT for work, even among coworkers doing the same kinds of tasks. In a study of about 17,000 entrepreneurs in Kenya who were all invited to use ChatGPT and shown how, women were still about 13% less likely to try it.

So the standard rollout, where IT sends out licenses and a link to a webinar, will probably reproduce the gap. There is some reason for optimism: Deloitte found US women adopted generative AI at about half the rate of men in 2023, and the gap was narrower in 2024. It is moving, but it is not closing on its own fast enough to ignore.

Where the gap opens inside a company

Women working in AI described specific barriers to the Guardian, as summarized here: referral-based hiring, few female mentors, 12-hour work cultures, and career breaks such as maternity leave. None of these are unique to AI. What AI does is speed things up, so a six-month absence or a missed informal invitation costs more than it used to.

In practice, I would watch four points. Hiring, where AI roles are often filled through networks before a posting goes up. Training, where optional sessions after hours favour people with fewer caregiving obligations. Project staffing, where the visible "let's try AI on this" pilots go to whoever volunteers loudest or already sits near the manager. And performance reviews, where AI-assisted output may simply look like higher productivity.

On that last point, honesty matters: the research for this piece found no studies on how performance reviews reward AI-assisted work or whether that creates bias. It is a reasonable concern, not an established fact. If one group uses AI tools more and reviews count raw output, the review system could quietly turn a usage gap into a rating gap. That is worth checking in your own data rather than assuming either way.

• Hiring: AI roles filled by referral before anyone else sees them
• Upskilling: optional, after-hours training that not everyone can attend
• Project access: AI pilots assigned informally to self-selected volunteers
• Reviews: output-based metrics that reward tool usage without asking about it
• Career breaks: people returning from leave to a toolset that has moved on

Europe and Türkiye are not ahead

Non-US employers should not take comfort in this being an American story. The European Institute for Gender Equality cites LinkedIn data showing only 16% of AI-skilled individuals in the EU are women, and notes that the gap widens as careers progress. On the exposure side, 39% of women's jobs in Europe and Central Asia are exposed to AI disruption, compared with 26% of men's.

Globally, the ILO found about 29% of female-dominated occupations are exposed to generative AI, against 16% of male-dominated ones. In the highest automation-risk categories, the figures are 16% and 3%. That is a large difference, and it lands on the administrative and service roles that many mid-size companies are now looking to automate first.

For Türkiye, the available data is thinner. Citing the Turkish Statistical Institute, one analysis puts women at less than 20% of the ICT workforce and less than 15% of engineering. An older 2019 report on WEF and LinkedIn data put women at 24% of AI experts in Turkey, slightly above a 22% global share. That figure is several years old and predates the current hiring wave, so I would not lean on it as evidence that Türkiye is doing better today.

Practical steps for managers and HR

The research does not hand employers a proven playbook. The searches behind this piece found no documented company upskilling program that measurably closed the gap. What it does tell you is that access alone fails, which points toward measurement and deliberate allocation rather than open invitations.

Randstad's Workmonitor 2025 found 44% of men ranked AI skilling as a top-three priority, compared with 36% of women. You can read that as a preference to respect, or as a sign that women see less payoff in the skill. Given the pay data above, I lean toward the second reading, which means the employer's job is to make the payoff visible and the path concrete.

The steps below are cheap to start and easy to drop if they do not help. The point is to treat AI adoption like any other process with a known failure mode: watch the numbers and intervene early.

• Measure tool usage by team and gender, the same way you track license adoption overall. If you do not look, you will not see the gap.
• Run AI training during working hours and make it part of the job, not a voluntary extra.
• Train in small groups by role, using real tasks from that team, so people learn on work they already know.
• Staff AI pilots deliberately. Rotate who leads them and write down why each person was chosen.
• Post internal AI roles openly with clear criteria before filling them through referrals.
• Give people returning from parental or long leave a short catch-up on tools and workflows that changed.
• Build paths out of highly exposed roles, for example from customer service into reviewing, configuring or quality-checking AI systems.
• Review your performance rubric: ask what was achieved and how, so AI use is credited consistently rather than only where managers happen to notice it.

What this looks like in a smaller firm

A 60-person company does not need a diversity task force to act on this. It needs one person who owns the AI rollout and a quarterly look at three numbers: who is using the tools, who attended training, and who got AI-related projects or new responsibilities. If those three lists skew heavily toward one group, you already know where to act.

What does not work, in my experience, is a single company-wide enthusiasm session followed by silence. The keen adopters get keener, everyone else assumes the tools are not meant for them, and six months later the informal "AI people" in the office are the same people who were already closest to leadership. Small firms are especially prone to this because so much runs on informal relationships.

Be modest about results. Since there is no evidence yet of a program that reliably closes the gap, treat whatever you try as an experiment, keep it light, and check whether the numbers move. If they do not, change the approach. The worst outcome is not trying something imperfect. It is discovering in two years' time that your pay gap widened while you were congratulating yourself on fast AI adoption.