There’s no shortage of optimism about AI in education — or anxiety, for that matter. Between promises of “personalised learning” and fears of “robot teachers,” it’s easy to forget that, in the end, classrooms run on people, not platforms. A recent study by Marta López-Costa and colleagues at the Universitat Oberta de Catalunya takes that human factor seriously. They asked what really predicts whether secondary school teachers start using AI in their work and what doesn’t.
Their analysis of more than 370 Catalan teachers shows something refreshingly concrete: it’s not having a STEM background or being especially tech-savvy that matters most. It’s knowing how AI actually works and being comfortable with data in general. Teachers who already used digital learning environments and could interpret student data were far more likely to adopt AI tools. In contrast, more specialised or technical forms of “data management” — such as privacy policies, data governance, and institutional charts — didn’t make much difference.
Perhaps most striking is that AI knowledge proved the strongest predictor of adoption. Teachers who had already experimented with generating content — images, lesson ideas, even snippets of code — were the ones integrating AI most actively. Knowing, not just hearing about, appears to be the bridge to using.
But the study also adds a note of caution. Teachers’ perceptions of AI — their ethical worries about bias, transparency or plagiarism — had a small but statistically significant adverse effect. In other words, concern still matters. Confidence in AI grows with understanding, yet unease remains a quiet counter-force.
Why this matters
If these findings hold more broadly, professional development may need a subtle rebalancing. Less emphasis on abstract “digital transformation” and more on practical AI literacy: learning how generative systems produce text or images, what counts as a reliable prompt, and where human judgement still trumps algorithmic output. Pair that with solid data-literacy training — how to read patterns in student performance data, how to tell noise from signal — and the chances of meaningful, ethical AI use rise sharply.
Equally important, the study reminds us that a teacher’s background in maths or science doesn’t guarantee readiness for AI. The ability to interpret data and to think pedagogically about technology cuts across subjects. AI adoption isn’t a STEM privilege; it’s a professional-learning challenge.
A balanced takeaway
The authors are careful not to overstate their results: their model explains about 30 per cent of the variance in AI adoption — decent but not decisive. And they used a cross-sectional design, so the cause-and-effect relationship remains an open question. Still, the pattern fits what many schools are already seeing in practice. Teachers adopt AI not because it’s imposed from above, but because they find concrete, pedagogically sound ways to use it.
So the message is both nuanced and straightforward: to make AI work in education, don’t start with AI. Start with teachers — their understanding of data, their curiosity, and their capacity to test, doubt, and adapt. Tools follow people, not the other way around.
Abstract of the study:
This study investigates the factors influencing the adoption of Artificial Intelligence (AI) by secondary school teachers in Catalonia. Using a Partial Least Squares Structural Equation Modelling (PLS-SEM) methodology, a conceptual model was analyzed that includes AI perception, AI knowledge, General data use, Applied data use, and STEM training as predictors of AI adoption. The results reveal that AI knowledge (β = .482, p < .001) and General data use (β = .288, p = .001) are the most significant and positive predictors of AI adoption. In contrast, AI perception shows a weak but statistically significant negative relationship (β = -.105, p = .022), while applied data use and STEM training do not present a significant direct effect. The model explains 30.5 % of the variance in AI adoption. These findings suggest that developing specific knowledge on how to use AI for content creation and competence in general data use is crucial to fostering AI adoption among secondary school teachers in the Catalan context. In addition, this explorative work provides the research community with evidence that key Data Literacy competencies significantly shape AI adoption.