Prompts, Textbooks, Maps: AI-Crafted Tools Grammar for SEN Students in an Italian High School

Artificial Intelligence (AI) is increasingly reshaping English language teaching through adaptive, learner-centred, and multimodal pedagogies, with particular relevance for inclusive education and students with Special Educational Needs. Within this framework, concerns regarding cognitive offloading, algorithmic hallucinations, and teacher overreliance call for balanced human–Artificial Intelligence collaboration. Grounded in Constructivism, Sociocultural Theory, Cognitive Load Theory, Adaptive Learning Theory, Self-Determination Theory, and the Technological Pedagogical Content Knowledge model, this study investigates how prompt engineering and generative Artificial Intelligence can support grammar instruction in a secondary school context. This article aims to explore the pedagogical potential of Artificial Intelligence-crafted tools for personalised English grammar instruction, examining how structured prompts, digital textbook resources, and multimodal outputs can support a student with special educational needs while preserving cognitive engagement and teacher mediation. A qualitative case study methodology was adopted in an Italian high school English classroom through a three-phase prompt engineering framework: (1) selection of instructional input resources, including textbook excerpts and cropped portable network graphics images; (2) structured prompt construction through role, context, instructions, and rule-based components; and (3) analysis of Artificial Intelligence outputs, including grammar exercises, conceptual maps, and interdisciplinary listening tasks. The study examined iterative prompt refinement and multimodal resource generation using open-access large language model systems and image-generating Artificial Intelligence tools. Particular attention was given to how prompt modifications influenced output complexity, accessibility, and pedagogical suitability for inclusive grammar teaching. Findings indicate that structured prompt engineering can support the generation of personalised, scaffolded learning materials aligned with inclusive didactic objectives. Artificial Intelligence-generated exercises and conceptual maps promoted simplification, visual support, and differentiated instruction, while prompt refinement improved output clarity and adaptability. Results also suggest that combining textual prompts with digital resources enhances specificity and supports multimodal learning. At the same time, the study highlights risks linked to excessive automation, including cognitive offloading and reduced critical engagement when teacher mediation is absent. Artificial Intelligence can function as a valuable didactic enhancer rather than a substitute for pedagogical expertise, particularly when integrated through reflective prompt design and theoretically grounded instructional practices. Effective implementation depends on balancing innovation with teacher judgment, ethical awareness, and student cognitive autonomy. The study suggests that prompt engineering may represent an emerging pedagogical competence for inclusive language teaching and calls for further research on the long-term cognitive and educational implications of Artificial Intelligence-supported instruction.