The rapid rise of generative artificial intelligence (GenAI) in education is reshaping how teachers design instruction, interact with learners, and understand their professional roles (OECD, 2023). Research highlights opportunities and challenges of GenAI  (Law, 2024; Zhai, 2024). GenAI tools can support differentiated learning, generate instructional materials, enhance student engagement, and reduce teachers’ workload by automating routine tasks (Hu et al., 2025; Chakraborty, 2024). At the same time, GenAI-generated outputs require pedagogical judgment, as issues of accuracy and ethical use complicate practice (OECD, 2023). These dual affordances illustrate the need to examine pedagogical, professional, and institutional conditions shaping teachers’ work. The aim of this study was to examine how pedagogical reasoning, teacher agency, and systemic conditions shape teachers’ integration of GenAI in school contexts. Accordingly, this study explored three research questions: (1) How do teachers integrate GenAI into their pedagogical practice? (2) How is teacher agency enacted within different depths and forms of integration? (3) How do institutional and systemic conditions enable or constrain GenAI adoption?

Techno-pedagogy is central to GenAI integration. The SAMR model (Puentedura, 2006, 2012) illustrates how technology use ranges from enhancement to transformation. Levy-Nadav et al. (2024) found most GenAI practices clustered in the middle SAMR levels, suggesting that meaningful integration depends on teachers’ pedagogical intentions and their ability to adapt GenAI outputs. Complementing SAMR, Dexter’s Educational Technology Integration and Implementation Principles (2005, 2023) emphasize value-added technology use, alignment between tools and learning goals, and the importance of supportive infrastructure. Collectively, these frameworks emphasize purposeful pedagogical adaptation. GenAI intensifies these demands, as teachers must evaluate content, refine prompts, and consider both pedagogical alignment and ethical concerns.

Teacher professional development (TPD) is another factor shaping GenAI integration. Research demonstrates that TPD enhances teachers’ competence and student outcomes (Harris & Sass, 2011). Specifically, in the context of GenAI, TPD must evolve to address emerging competencies such as prompt refinement, evaluation of GenAI outputs, and responsible implementation in diverse subject areas. Recent work shows that structured exploration and reflection support informed GenAI use (Ding et al., 2024). These findings reinforce the importance of TPD programs that cultivate teachers’ capacity to integrate GenAI thoughtfully.

Teacher agency is a further lens through which to understand GenAI adoption. Agency encompasses autonomy, innovation, and ownership over professional growth (Calvert, 2016; Imants, 2020). Teachers exercise agency when they critically evaluate when and how GenAI should be used, adapt GenAI-generated materials to student needs, and initiate new instructional strategies. Studies indicate that teachers’ perceptions of GenAI vary widely; some embrace experimentation, while others proceed cautiously, concerned about accuracy, ethics, or alignment with learning goals (Law, 2024; Zhai, 2024). These variations underscore that integration depends not only on technical skills but on teachers’ professional judgment, sense of responsibility, and confidence navigating new technologies.

Finally, systemic and institutional conditions influence GenAI implementation. Infrastructure, policy clarity, administrative support, and ethical guidelines shape the extent to which teachers engage with GenAI (OECD, 2023). Without reliable access, clear expectations, or supportive leadership, even highly motivated teachers encounter barriers that restrict experimentation or deeper innovation. Concerns about academic integrity, and responsible use further highlight the need for coherent institutional frameworks (UNESCO, 2023; Roe & Perkins, 2024). Conversely, environments that encourage collaboration enable sustainable and pedagogical adoption (Imants, 2020).

The literature suggests that GenAI integration emerges through the interaction of three interrelated dimensions: pedagogicalreasoning informed by techno-pedagogical frameworks, teacher agency as expressed through professional judgment and innovation, and system-level conditions that support or constrain practice. Understanding these dimensions provides a conceptual foundation for examining how teachers navigate GenAI in real contexts and identifies the factors that shape meaningful, ethical, and sustainable integration.

This qualitative study employed a design which triangulated interviews, observations, and GenAI-integrated teaching artifacts to examine teachers’ GenAI integration across school contexts. This design enabled examining in depth the pedagogical, agentic, and systemic factors shaping teachers’ integration of GenAI across school contexts. Participants: Seventeen in-service teachers from a GenAI-focused TPD participated in the study. They represented a range of subject areas and teaching experience, enabling an examination of GenAI use across varied pedagogical and institutional realities. Instruments: Data were collected from three sources: semi-structured interviews, observations of professional development sessions, and 91 GenAI-integrated teaching artifacts produced during and after the training. Individual interviews explored teachers’ reasoning behind adopting GenAI, perceived pedagogical affordances and challenges, experiences of autonomy or constraint, and views on institutional or policy factors shaping their decisions. Approximately ten hours of observations captured how teachers learned with and about GenAI in collaborative settings, including how they experimented with prompt crafting, interpreted GenAI outputs, supported peers, and responded to the instructional guidance provided during the training. The teaching artifacts included lesson plans, adapted materials, assessments, and student-facing tasks generated or revised using GenAI, offering evidence of integration depth and the practical implications of pedagogical or systemic barriers. Analysis: Data were analyzed using thematic analysis. Coding combined deductive categories derived from techno-pedagogical principles (e.g., alignment with learning goals, value-added use), expressions of teacher agency (autonomy, reflection, collaboration, innovation), and system-level factors (infrastructure, policy clarity, administrative support), alongside emerging inductive codes. The aim was to identify patterns that explained variation in GenAI uptake across teachers and contexts. Results: Three findings emerged. First, teachers’ GenAI practices were influenced by their pedagogical decision-making: those who aligned tool use with learning goals demonstrated deeper and more adaptive integration, whereas others remained at surface-level use. Second, teacher agency played a decisive role. Teachers who felt confident to experiment, reflect, and collaborate integrated GenAI more creatively and critically; those facing uncertainty or restrictive norms adopted it minimally. Finally, systemic conditions shaped the boundaries of what teachers could implement. Clear policies, supportive leadership, and reliable infrastructure enabled sustained engagement, while ambiguity or constraints limited integration even among motivated teachers. Together, these findings illuminate how pedagogy, agency, and context interact to shape GenAI school adoption.

The findings demonstrate that teachers’ integration of GenAI is shaped by the interaction of pedagogical reasoning, teacher agency, and systemic conditions. Teachers who aligned GenAI use with learning goals, refined outputs, and adapted materials for diverse learners engaged in deeper integration, highlighting that GenAI adoption is fundamentally a pedagogical process rather than a technical one. At the same time, agency played a pivotal role: teachers who felt confident to experiment, collaborate, and influence the professional development process used GenAI more creatively and critically, reinforcing the centrality of reflective judgment and innovation when working with emerging technologies. System-level factors also significantly structured what teachers were able to enact. Clear expectations, ethical guidelines, supportive leadership, and reliable infrastructure enabled sustained and thoughtful engagement, whereas ambiguity or restrictive norms limited integration even among motivated teachers. Beyond its practical implications, the study offers a theoretical contribution by proposing that GenAI integration should be conceptualized not as movement along a single pedagogical scale but as a multi-dimensional process shaped by the alignment of three interdependent dimensions: The pedagogical dimension emphasizes expanding GenAI training for purposeful prompt design and higher-level integration. The agency dimension includes collaborative learning, and opportunities to influence the design of TPDs, both of which expand teachers’ autonomy and reflective practice. The systemic dimension encompasses infrastructure investment, efficiency benefits and policy development that include ethical guidelines enabling or constraining sustainable GenAI use. This perspective extends existing techno-pedagogical models by explaining how higher-level GenAI use depends on the interaction between teachers’ judgment and contextual enablers, rather than on technological affordances alone. As a limitation, the qualitative design captures depth across contexts but does not allow claims about generalizability of findings. Future research could build on this model using mixed-methods to examine how these dimensions interact at scale.

Calvert, L. (2016). The power of teacher agency. The Learning Professional, 37(2), 51. Chakraborty, S. (2024). Generative AI in modern education society. arXiv. https://arxiv.org/abs/2412.08666. Dexter, S. (2005). Principles to guide the integration and implementation of educational technology. In M. Khosrow-Pour (Ed.), Encyclopedia of Information Science and Technology (1st ed., pp. 2303-2307). IGI Global. https://doi.org/10.4018/978-1-59140-553-5.CH406 Dexter, S. (2023). Developing faculty EdTech instructional decision-making competence with principles for the integration of EdTech. Education Tech Research Dev, 71, 163–179. https://doi.org/10.1007/s11423-023-10198-0 Ding, A.-C. E., Shi, L., Yang, H., & Choi, I. (2024). Enhancing teacher AI literacy and integration through different types of cases in teacher professional development. Computers and Education Open, 6, 100178. Harris, D. N., & Sass, T. R. (2011). Teacher training, teacher quality and student achievement. Journal of Public Economics, 95(7-8), 798-812. https://doi.org/10.1016/j.jpubeco.2010.11.009 Hu, X., Xu, S., Tong, R., & Graesser, A. (2025). Generative AI in Education: From Foundational Insights to the Socratic Playground for Learning. arXiv. https://arxiv.org/abs/2501.06682 Imants, J., & Van der Wal, M. M. (2020). A model of teacher agency in professional development and school reform. Journal of Curriculum Studies, 52(1), 1-14. https://doi.org/10.1080/00220272.2019.1604809 Law, L. (2024). Application of generative artificial intelligence (GenAI) in language teaching and learning: A scoping literature review. Computers and Education Open, 6, 100174. https://doi.org/10.1016/j.caeo.2024.100174 Levy-Nadav, L., et al. (2025). Digital Competencies for Effective GenAI Use in Secondary Schools: A Longitudinal Exploration of Teachers' Perspectives and Classroom Practices. Journal of Computer-Assisted Learning. JCAL_EV_JCAL70123 OECD. (2023). Generative AI in the classroom: From hype to reality? OECD Schools+. https://one.oecd.org/document/EDU/EDPC(2023)11/en/pdf Puentedura, R. (2012). The SAMR model: Six exemplars. Retrieved November 15, 2023 from http://www.hippasus.com/rrpweblog/archives/2012/08/14/SAMR_SixExempl ars.pdf Puentedura, R. (2006). Transformation, technology, and education. Retrieved from http://hippasus.com/resources/tte Roe, J., & Perkins, M. (2024). Generative AI and agency in education: A critical scoping review and thematic analysis. arXiv. https://arxiv.org/abs/2411.00631 Shamir‐Inbal, T., & Blau, I. (2021). Characteristics of pedagogical change in integrating digital collaborative learning and their sustainability in a school culture: e‐CSAMR framework. Journal of Computer Assisted Learning, 37(3), 825-838.‏ http://dx.doi.org.elib.openu.ac.il/10.1111/jcal.12526 UNESCO. (2023). AI and education: Guidance for policy-makers. Retrieved May 17, 2024, from https://unesdoc.unesco.org/ark:/48223/pf0000386162 Zhai, X. (2024). Transforming teachers' roles and agencies in the era of generative AI: Perceptions, acceptance, knowledge, and practices. arXiv. https://arxiv.org/abs/2410.03018