Research from the 2026 ICGS Global Action Research Collaborative on Girls' Education (GARC) explores the use of AI in girls’ schools bring together classroom studies from Year 1 to Year 12, spanning subjects from math to cybersecurity to visual art. They show that teachers in girls’ school can leverage AI to build something powerful: real confidence, a willingness to take risks, and a stronger sense of ownership over learning in STEM, writing, and leadership. Four themes emerged in their work.

1. AI as a Psychological Sanctuary for Risk-Taking

Across age groups, the research keeps landing on the same insight: AI works best when it lowers the emotional stakes of being wrong.

In Lauren Parker's Year 6 maths study (Notting Hill and Ealing High School, UK), girls described the dread of a hard problem — feeling "sad and worried because everyone else looks like they can do it." Stripping out timers, scores, and leaderboards in favor of a non-judgmental AI platform shifted students toward "willingness to have a go." One called the work "challenging at times but… made me feel really independent."

In Kerry Daud's Year 8 digital design study (St Margaret's Anglican Girls School, Australia), an anonymised AI creative space gave perfectionist students permission to be messy. Students who once asked "how am I supposed to know if it is right?" ended up calling AI "your creative partner, not a replacement for your own thinking."

By Year 12, the stakes shift to exam pressure, but the sanctuary effect holds. In Jayne Schinckel's psychology study (St Hilda's School, Australia), students using "Bloom AI" as a Socratic tutor said: "It is supportive and given me more confidence, and I'm not as stressed knowing I have something that can help me when I get stuck."

2. From Answer Machines to Thinking Partners

The second theme is a warning label: AI that gives answers can undercut the thinking it's meant to support. The reports that worked hardest to avoid this found that friction, not fluency, builds confidence.

In Susan Jackson's Year 8 maths study (Diocesan School for Girls, New Zealand), an AI coach was deliberately restricted from giving solutions. Students reported it "made me more likely to try challenging problems" and made it "less easy to give up."

In Jo Oreo's Year 9 science study (Ruyton Girls' School, Australia), a "co-thinking" framework pushed students to record their own thinking before consulting AI. One student reflected: "I used to think AI was mostly there to be used as a source of information, and now I think it can help me transform my own ideas… whilst using my thinking."

By Year 12, this becomes explicit critical evaluation. In Tina Huang's English study (All Hallows' School, Australia), students moved from treating AI as an oracle to interrogating it: "you have to critique AI because it's not, obviously, human, and it doesn't always know 100%."

3. Building Critical Agency Through Adversarial and Ethical Auditing

The third theme is the most striking departure from "AI as helper": several reports had girls test AI's flaws and biases as a form of empowerment.

Thomas Heverin's cybersecurity study (The Baldwin School, USA) had students attempt "prompt injection,” that is prompts meant to break AI models' rules, while using AI chatbots. Confidence shifted dramatically — from "I don't feel at all confident" to "I feel very confident now that I can create successful prompt injections" — alongside new ethical unease: "what if someone with malicious intent is using this stuff?"

Sydney Costa's Year 8 GALS Series study (Girls Athletic Leadership School, USA) had students audit AI image generators for bias. Findings were blunt: "AI only generated a white woman when we said to have a person with colored hair… basically no one's AI made a different skin color."

At the youngest end, Hugh Earlam's Year 6 study on AI, social media, and discernment (Seymour College, Australia) made critical thinking almost reflexive: "Every time I look at something, there's a voice that says this could be AI." That awareness came with emotional weight: "I know it's not real, but I still feel sad."

4. The Human-in-the-Loop Advantage

The final theme connects every age group: AI tools succeed or fail based on students' relationship with the teacher behind them.

In Deane Valodimos's Year 1 numeracy study (Genazzano FCJ College, Australia), the teacher — not students — was the primary AI user, leveraging Perplexity to build vocabulary-rich lessons. The payoff showed up in how six-year-olds described themselves: "I am a mathematician because I'm good at subtraction."

In Mary Jane Kennedy's Grade 9 English study "Ceci," (Marymount School of New York, USA), students drew a firm line between chatbot and teacher: "Teachers know what they're looking for. Ceci [the AI chatbot] doesn't know exactly what the teacher's looking for."

In Perri Winter's Year 12 visual arts study (Presbyterian Ladies' College, Australia), trust in the teacher was the whole reason students engaged honestly: "probably more likely to use it… because I know you're checking whether we're on track." That trust had limits: "The feedback part of AI is not trustworthy for marking of exams."

In Summary

What these reports show, taken together, is that AI can genuinely build skills such as confidence, judgment, resilience, agency when it's used thoughtfully. Teachers need to stay visibly in the loop and encourage students to use AI as scaffolding rather than a shortcut. Handled that way, a six-year-old can start calling herself "a mathematician," and a seventeen-year-old can learn to question an AI model instead of trusting it outright. As Susan Jackson's research puts it, the goal is simply to "give the bird wings.": design the thinking first, and decide where AI fits second.

For more in-depth insights, explore the full action research reports from the GARC 2025–2026 cohort at research.girlsschools.org.