Her Impact AI · Annual Report
The State of Women in AI 2026
Women closed the AI usage gap and now lead on AI strategy. Then they hit a wall. The 2026 data shows the constraint is no longer whether women use AI. It is how their use gets read.
Published July 2026 · Her Impact AI · herimpactai.com
The headline finding
Usage gap closed. Recognition gap open.
In 2026, women use AI at nearly the same rate as men and are more likely to bet on it for growth. What has not moved is how that work is read. Identical output is scored as competence from a man and as a shortcut from a woman. The binding constraint has shifted from access to recognition, and that is a strategy problem, not a skills problem.
How to read this report
Four dimensions, one system
Her Impact AI reads this data through the Impact Architecture: the four dimensions that decide whether technical fluency converts into influence. They are not stages. They operate together, and at any moment one is the binding constraint. The 2026 evidence maps cleanly onto all four, which is the real story of the year. The frame throughout is credibility asymmetry, not a confidence gap. The women in this data are not less capable or less willing. They are read differently.
Recognized expertise
Whether your judgment is read as the one worth having. In 2026 this is where the competence penalty lands.
Distribution and credit
Whether your work travels and is attributed to you. The praise and encouragement gaps live here.
Proximity to decisions
Whether you are in the rooms setting the AI agenda. Women have the strategic read; the question is the room.
Durable trust
Whether you are trusted over time. In 2026 this is taxed by the perception that using AI is cheating.
The setup
The usage gap is no longer the story
For years the headline was adoption: women used AI less, so the fix was to get them using it more. That gap has essentially closed. Pew’s February 2026 survey of 5,119 US adults puts overall chatbot use at 50% for men and 47% for women, down from an 11 point gap in 2024. At work, Lean In finds 78% of men and 73% of women have used AI. On the single most common tool, ChatGPT, use is identical at 44%.
Sources: Pew Research Center (overall use, ChatGPT), Lean In (use at work), 2026.
Two gaps remain, and they matter for what follows. Women report using the more specialized tools less, where AI does its most differentiated work: Gemini 20% versus 29%, Copilot 13% versus 22%, Claude 4% versus 9%. And women report lower confidence, 15% versus 22%. Read on its own, that looks like a skills story. Read against the recognition data below, it looks like a rational response to a system that penalizes women more sharply for the same visible AI use.
Dimension 01 · Authority
The competence penalty
Same work. Discounted read.
The clearest finding of the year is that using AI carries a competence penalty, and the penalty is not evenly distributed. In a controlled study published in Harvard Business Review, 1,026 engineers reviewed identical Python code. The only thing that changed was whether the code was labeled as AI assisted. AI users were rated 9% lower on competence for the same work. For women that penalty roughly doubled.
Source: Acar, Gai, Tu & Hou, reported in Harvard Business Review, 2025. Same code, ratings varied only by AI-use label and coder gender.
A second 2026 study by researcher Zehra Chatoo isolated the mechanism. One thousand UK adults evaluated identical AI assisted resumes where only the name changed, Emily Clarke or James Clark. Evaluators were 22% more likely to doubt the woman’s trustworthiness. Among Gen Z male evaluators the split was 97% rating James as strong versus 76% for Emily, a 21 point gap on identical material. Her one line names the asymmetry precisely.
This is why the confidence gap, 15% of women versus 22% of men reporting high confidence, is better read as a symptom than a cause. Authority is competence made visible and trusted by others. When the same visible signal is discounted for you, caution is not a deficit. It is accurate pattern recognition. The move is not to hide the tool or collect another credential. It is to make judgment the visible part, so recognition attaches to the reasoning only you could have supplied.
Read the Authority dimension →Dimension 02 · Reach
The credit gap
The work travels less, and the credit even less.
If Authority is about how work is read, Reach is about whether it is seen and attributed at all. Here the 2026 data is blunt. Lean In’s March survey finds men are 27% more likely to be openly praised for AI use at work and 23% more likely to be encouraged by a manager to use it. The same behavior produces visible recognition for one group and silence for the other. Over a year, that silence compounds into a story about who is naturally the AI-forward operator, and that story shapes who gets the next mandate.
Source: Lean In, 1,015 US adults, March 2026.
The strategic reading is not to wait for the credit to arrive. It is to build distribution deliberately, so the record of what you did exists outside the room and outside anyone’s memory. Reach is not volume or self-promotion. It is a citable trail that assigns credit before the question is ever asked.
Read the Reach dimension →Dimension 03 · Access
The vision-to-room gap
Women have the strategic read. The question is proximity.
The most overlooked finding of 2026 is that senior women are not the cautious ones on AI strategy. They are the ambitious ones. AlixPartners surveyed 3,000 C-suite and senior executives for its 2026 Disruption Index. Women were more likely than men to treat AI as a growth instrument rather than a cost lever, more likely to be extremely optimistic, and more likely to be running enterprise-wide agentic rollouts.
Source: 2026 AlixPartners Disruption Index, 3,000 senior executives worldwide.
This reframes Access entirely. The issue is not conviction or readiness. Women leaders are reading the deployment landscape with more precision, not less: fewer of them expect near-term AI layoffs, 61% versus 71% for men, because they are closer to where AI is actually working. The strategic risk is that this sharper read stays inside their own heads. Access is proximity to the rooms where the AI agenda gets set. When you already hold the better thesis, the highest-leverage move is engineering your way into the decision, not refining the thesis further.
Read the Access dimension →Dimension 04 · Reputation
The integrity tax
Trust is the asset being charged.
Reputation is the difference between being known and being trusted, and it is where the year’s data turns from competence to character. Lean In finds women are 32% more likely than men to worry about being perceived as cheating when they use AI, 29% versus 22%. Chatoo’s study shows why the worry is rational: the doubt attached to a woman’s AI use was about trustworthiness, not ability. The tax is levied on trust, the one asset that compounds.
The rational move in the moment is the expensive move over time.
Hiding AI use protects against today’s suspicion. It also forfeits the visible judgment that builds a reputation, and it cedes the AI-forward narrative to people who pay no penalty for claiming it. Quiet is not neutral. In a year when AI made everyone’s output look similar, visible judgment became the scarce, trusted signal.
The response is not to over-disclose or to perform certainty. It is to build a reputation deliberately, on consistency and kept, witnessed commitments, so that trust accrues faster than the tax can erode it. A reputation built on purpose is the one asset in this report that no one can discount by relabeling your work.
Read the Reputation dimension →Why it matters now
Highest exposure, lowest representation, discounted work
These gaps would matter in any year. They matter more in 2026 because of where women sit relative to the disruption. In March 2026 the International Labour Organization found that 29% of female dominated occupations are exposed to generative AI, versus 16% of male dominated ones. In the highest risk category the split is starker: 16% versus 3%. Across 88% of countries analyzed, women’s work was more exposed than men’s.
Put the year together and the position is specific: women are the most exposed to AI’s disruption, the least represented in the workforce building it, and the most penalized for using it visibly, while holding the more ambitious strategic read on what it is for. That is not a story about catching up. It is a story about a narrow window in which the prototype of who is credible in AI is still being written.
The strategic response
What the data tells women to do differently
If the constraint were adoption, the answer would be more usage. If it were skill, the answer would be more training. The 2026 data points elsewhere. The constraint is recognition, and recognition is built, not awarded. Across the four dimensions the moves are consistent: make judgment visible rather than hiding process, build a citable record so credit does not depend on memory, engineer proximity to the decisions you already read well, and construct trust deliberately so it outpaces the integrity tax.
None of that requires being louder. It requires being deliberate about the part of your work that no relabeling can discount: the reasoning only you could supply, made legible on the record. That is the difference between using AI and being recognized for the judgment behind it.
Find your primary constraint
The Impact Architecture diagnostic maps where you are strong and where you are stuck across all four dimensions, and names the one to work first. It takes about three minutes.
Take the diagnosticMethodology & sources
The studies behind this report
Every figure in this report is drawn from published 2025 to 2026 research. Where a study reported raw percentages, they are quoted directly. Comparative bars show the two groups as reported in the source.
- Pew Research Center, “The gender gap in AI” (June 17, 2026). Survey of 5,119 US adults, February 17 to 23, 2026. Overall and daily chatbot use, tool-level usage, confidence, productivity, and views on AI’s pace and impact.
- Lean In, “Women Use AI Less at Work and Get Less Credit” (2026). Nationally representative survey of 1,015 US adults, March 2 to 6, 2026, margin of error plus or minus 3.1 points. Workplace use, praise, manager encouragement, and the perception of being seen as cheating.
- Acar, Gai, Tu & Hou, “The Hidden Penalty of Using AI at Work,” Harvard Business Review (2025). Controlled evaluation of identical code by 1,026 engineers, with a supporting dataset of 28,698 software engineers. The competence penalty and its heavier weight on women and older workers.
- Zehra Chatoo, AI resume perception study (April 2026). 1,000 UK adults evaluating identical AI assisted resumes with only the candidate name varied. Trustworthiness doubt and the effort-versus-integrity framing.
- 2026 AlixPartners Disruption Index. Survey of 3,000 C-suite and senior executives worldwide. Gender differences in AI as growth versus cost, optimism, agentic rollout, and layoff expectations.
- International Labour Organization (March 5, 2026). Analysis of generative AI occupational exposure by gender across countries, plus women’s share of the global AI workforce.
A note on framing. Her Impact AI reports these gaps as credibility asymmetry, not a confidence or competence deficit in women. The controlled studies hold work quality constant and vary only perception, which locates the gap in how the work is read, not in the work itself.