Teachers Report Sharper AI Awareness After University-Wide Training Push

When a Spanish university decided to treat generative artificial intelligence not as a threat to be policed but as a professional competency to be taught, it set in motion one of the first longitudinal experiments in faculty attitudes toward the technology. A new pilot study from Universidad Alfonso X el Sabio (UAX) in Madrid, published in the Journal of New Approaches in Educational Research, suggests that a coordinated institutional training plan can measurably shift how university teachers understand, value, and intend to use generative AI in their work. The findings arrive at a moment when higher education institutions worldwide are scrambling to formulate responses to tools such as ChatGPT, often without clear evidence about what kinds of intervention actually change teacher behavior.

The research team, led by Maria Dolores Vivas-Urias, Cristina Obispo-Diaz, and Maria Auxiliadora Ruiz-Rosillo, tracked 180 university teachers across two consecutive academic years, 2023–2024 and 2024–2025, using a pre-experimental one-group pre-test and post-test design. Rather than comparing trained teachers against an untrained control group, the study measured the same cohort twice: once before and once after the rollout of the university’s GenAI Institutional Plan 2023–2025. This design allowed the researchers to capture within-person change on three core dimensions: general awareness of generative AI, beliefs about its impact on professions, and the perceived importance of learning AI tools for integration into student training.

The headline result is striking in its magnitude. Teachers’ self-reported awareness of generative AI rose by an average of 2.24 points on a ten-point scale, climbing from 4.82 to 7.06, a difference the authors characterize as clear and statistically significant. Perceptions followed a parallel trajectory. By the second year, 80.9 percent of surveyed teachers believed generative AI would transform the way they teach, an increase of 3.76 percentage points. Belief in the technology’s impact on students’ professional futures also rose, with the overall mean score increasing from 7.97 to 8.67 out of 10. Perhaps most tellingly for institutional planners, the perceived importance of learning AI tools for integration into student training reached 8.63 in 2024–2025, a statistically significant gain of 0.56 points over the previous year.

Beneath these averages lies a structural insight that may prove the study’s most consequential contribution: awareness, perceived professional impact, and the motivation to learn form an interlocking psychological chain. Correlation analysis revealed a strong positive relationship of r = 0.718 between teachers’ perceptions of generative AI’s impact on students’ professional futures and the importance they attached to learning AI tools. A more moderate correlation of r = 0.316 linked overall AI awareness to perceived professional impact. In practical terms, teachers who understand the technology better are more convinced it will reshape their fields, and that conviction drives their appetite for training. The authors interpret this through established frameworks such as the Technology Acceptance Model, in which performance expectations, the belief that adopting a technology improves outcomes, fuel adoption.

The institutional machinery behind these shifts deserves scrutiny. UAX launched its GenAI Institutional Plan in the 2023–2024 academic year, establishing milestones for both short- and medium-term transformation. The centerpiece for faculty was a compulsory training program, the Level II Teaching Certification in Generative Artificial Intelligence, deliberately aligned with UNESCO’s AI Competency Framework for Teachers across its five dimensions: a human-centred mindset, ethics of AI, AI foundations and applications, AI for professional learning, and AI pedagogy. The certification combined asynchronous online micro-courses introducing fundamentals, ethical use, data privacy, and academic integrity with face-to-face or synchronous workshops in which teachers practiced prompt design, generated learning activities mapped to Bloom’s Taxonomy, and built assessment strategies and personalized educational materials.

Participation figures point to genuine appetite rather than mere compliance. Of 1,300 invited teachers, 65 percent engaged in at least one training activity, and 428 completed the full certification. Satisfaction ratings were high across the board: the certification scored 8.6 for course level and 8.5 for usefulness for professional development, while the shorter two-hour practical workshops scored even higher, reaching 9.1 and 9.2 on usefulness and recommendation respectively. Notably, 72 percent of participants reported they had been able to apply what they learned in their personal and professional lives, and the training’s culminating activity required each teacher to submit a concrete proposal for integrating AI into their own teaching practice. That contrasts with prior research in which educators reported strong general technology training yet still felt unprepared to work generative AI into their classrooms.

Behavioral indicators tracked alongside perceptions tell a consistent story. The share of teachers reporting no use of AI tools collapsed from 54.38 percent in 2023–2024 to 10.42 percent a year later, and use of AI for teaching purposes jumped from 16.59 percent to 43.40 percent of responses. ChatGPT remained the dominant tool, cited by 89.14 percent of users in 2024–2025, but Copilot usage rose sharply, from 7.57 percent of responses to 28.99 percent, likely aided by the institution’s provision of Microsoft Copilot through university Office 365 accounts. Nearly half of the sample, 45.73 percent, used both tools concurrently. Because 72.6 percent of participating teachers held part-time contracts and worked outside the university, the faster uptake in teaching contexts relative to professional practice suggests that institutional training, rather than external professional exposure, was the decisive accelerant.

The study is candid about its constraints. The absence of a control group means the changes cannot be causally attributed to the training program alone; broader cultural momentum around generative AI between 2023 and 2025 undoubtedly contributed. The voluntary response rate was low, at 16.6 percent of the overlapping faculty population, raising the possibility of self-selection bias in which the most motivated teachers dominated the sample. The measurement instrument itself, a three-item one-dimensional perception scale, showed acceptable reliability with a Cronbach’s alpha of 0.72 and omega of 0.76, but the authors acknowledge that the saturated factor model limits interpretive weight of the confirmatory fit indices, and the modest Kaiser–Meyer–Olkin value of 0.59 indicates restricted factorability that future work should address. Reliance solely on teacher self-report, without observational evidence of changed classroom practice, further narrows what the data can establish.

Qualitative responses flesh out where teachers want to go next. An inductive thematic analysis of open-ended answers from 67 respondents identified seven broad categories and 29 specific training needs, clustered around pedagogical integration of generative AI, personalized content creation, automated assessment, ethics and responsible use, contextualized technical development, and certified continuing education. The authors note that most of these demands map onto the existing certification program, but more technical requests, such as discipline-specific tool development, will require specialized professional development beyond the current generalist offer. Faculties of Education Sciences and Technology and Business led awareness scores in both years, while significant pre-post gains appeared across Healthcare Sciences, Medicine, Dentistry, and Veterinary Medicine, indicating that the training effect penetrated fields far removed from computer science.

For the global higher education sector, the study offers a rare data point in a debate dominated by policy documents rather than measured outcomes. Recent international analyses have catalogued how universities are issuing guidelines, redesigning assessment, and drafting ethical codes, but few have quantified how such institutional efforts move the people on the front lines. The UAX results support a systemic, ethical, and sustainable strategy in which AI literacy is treated as an ongoing faculty competency rather than a one-off compliance exercise, and in which generative AI is embedded directly into curricula to close the gap between university education and labor market demands. The authors recommend that future research measure the transfer of training into actual teaching practice, incorporate student perspectives, and complement self-report with behavioral evidence. If replicated with stronger designs, the core message, that well-structured institutional training can convert teacher anxiety about AI into informed professional momentum, could reshape how universities worldwide approach the fastest-moving educational transformation of the decade.

Subject of Research: University teachers' perceptions of generative artificial intelligence following an institutional training plan in higher education

Article Title: Analysis of teachers’ perceptions of the impact of Generative Artificial Intelligence in higher education

Article References: Vivas-Urias, M. D., Obispo-Díaz, C., & Ruiz-Rosillo, M. A. (2026). Analysis of teachers’ perceptions of the impact of Generative Artificial Intelligence in higher education. Journal of New Approaches in Educational Research, 15(1), Article 12. https://doi.org/10.1007/s44322-026-00060-5

Image Credits: AI Generated

DOI: 10.1007/s44322-026-00060-5

Keywords: generative artificial intelligence, higher education, teacher training, AI literacy, faculty perceptions, institutional policy, UNESCO AI competency framework, ChatGPT, educational innovation, pre-test post-test study, teacher professional development, Spain