As digital technologies proliferate in student's lives, teachers are increasingly called upon to ensure that their use remain inclusive, equitable, and pedagogically sound. The spread  of generative AI - particularly text-to-text applications - has already sparked debate around writing integrity and originality (Fleckenstein et al., 2024; Barrett & Pack, 20223). Less explored, however, are text-to-image AI tools (e.g., DALL-E 3), which can serve as visual feedback to enrich writing instruction. This paper addresses the ways in which AI text-to-image technologies can foster inclusive writing pedagogy, by providing real-time, multimodal feedback that may benefit diverse learners—including those with limited language proficiency or learning disabilities (Liu et al., 2024).

Drawing on process-based writing theories (Hayes & Flower, 1980) and socio-constructivist principles of multiliteracy (Kress, 2003; Unsworth, 2006), the project emerged from a school-university collaboration in the metropolitan area of Turin (Italy). Three lower secondary teachers—two from Grade 6 (first year) and one from Grade 7 (second year)—participated in a research-based professional development program focused on: emphasizing explicit instruction in which AI text-to-image tools were integrated, as a scaffold for enriching and refining text and as a multimodal support. 

The core activity here presented enabled students to submit parts of their written drafts as prompts to DALL-E 3. The AI would generate an image reflecting the text’s descriptive details. Students then examined how closely the generated visuals aligned with their intended meaning, thanks to the mediation of the teachers. Any discrepancies (e.g., missing character traits, unclear setting details) became catalysts for revision. Teachers mediated this process by guiding students in adjusting their written descriptions or prompts, thus fostering critical thinking, vocabulary expansion, and deeper engagement with the writing task.

This inclusive, AI-assisted approach centers on two main research questions:

1. Is this tool effective for favoring critical thinking and revision processes in students?

2. And what for word fluency and richness of descriptions?

Rather than offloading the writing process to AI, this approach situates the technology as a stimulus for critical thinking and revision (Guilfors, 1967; Anderson & Krathwohl, 2001; Ennis, 2018; Hayes & Flower, 1980). By visually illustrating the “mental images” that student texts generate, AI and the mediation of the teacher can increase awareness of how written language can be improved to convey richer, more precise meaning.

A mixed-methods design was used to capture both the qualitative and quantitative dimensions of this classroom-based intervention. The study was divided in two phases: 1. Professional Development: Three lower secondary teachers completed a course on writing pedagogy and AI text-to-image applications. Based on the infusion model (Calonghi & Boncori, 2006; Trinchero, 2022), the teachers learned to incorporate generative AI (Su & Yang, 2023) within their teaching for developing higher order cognitive skills. The rsearcher and teachers together developed lesson plans focusing on genre-based writing (fantasy in Grade 6; mystery/giallo in Grade 7). Most lessons incorporated a feedback loop where the student’s draft would be transformed into an AI-generated image. 2. In class intervention: From February to May 2024, each teacher implemented the AI-enhanced writing activities with around 20 students per class. The AI (DALL-E 3) was used collectively (one student’s text at a time) to facilitate explicit scaffolding, with the teacher modeling prompt refinement and critical reflection. While teachers implemented in their classrooms the innovative approach, the researcher conducted participant observations, focusing on how learners of varying abilities engaged with the AI tool. Particular attention was paid to how teachers differentiated instruction or supported students with fragile writing skills. Some of the students’ original and revised drafts were collected. Changes in descriptive richness, coherence, and overall clarity were tracked. After the work ended, semi-structured interviews explored teachers' perceptions of AI’s utility, its inclusive potential, and any ethical or logistical challenges (e.g., time constraints, digital equity).

Visual aids appeared beneficial for students with diverse linguistic or learning needs - especially those who struggled with abstract descriptions. The AI-generated images served as a “mirror” reflecting the text’s strengths and omissions, leading to reflection and correction, thanks also to the scaffolding the teachers provided. In fact, teacher mediation was critical. Without supportive mediation, some students were initially unsure how to interpret or act upon the AI’s often imperfect visual output (e.g., misrepresentations of hair color or character attributes). Ethical considerations emerged around potential biases in image generation and whether students might rely too heavily on AI for feedback. Teachers addressed these concerns through collective discussions on the topic, emphasizing that AI outputs do not replace a writer’s imaginative process but rather spark critical thinking and deeper elaboration. For the final stage of data analysis, we aim to compare pre-intervention and post-intervention writing samples to assess growth in fluency of words and richness of descriptions. Ultimately, this pilot suggests that text-to-image AI can be integrated into writing instruction as an inclusive tool if carefully introduced and supported in the process, opening new frontiers for digital media–based feedback. However, the findings are not definitive and ongoing discussions around ethics, equity, and teacher training.

Anderson, L. W., & Krathwohl, D. R. (Eds.) (2001). A taxonomy for learning, teaching, and assessing: A revision of Bloom’s taxonomy of educational objectives. Allyn & Bacon. Andrzejczak, N., Trainin, G., & Poldberg, M. (2005). From image to text: Using images in the writing process. International Journal of Education & the Arts, 6(12), 1–17. Azevedo, R., et al. (2011). Adaptive content and process scaffolding: A key to facilitating students’ self-regulated learning with hypermedia. Psychological Test and Assessment Modeling, 53(1), 106–130. Barrett, A., & Pack, A. (2023). Not quite eye to AI: Student and teacher perspectives on the use of generative artificial intelligence in the writing process. International Journal of Educational Technology in Higher Education, 20(1), 59. Boscolo, P., & Zuin, E. (2015). Come scrivono gli adolescenti. Bologna: Il Mulino. Calonghi, L., & Boncori, L. (2006). Guida per la correzione dei temi. LAS. Chicho, K. Z. H., & Zrary, M. O. H. (2022). Using Visual Media for Improving Writing Skills. Canadian Journal of Language and Literature Studies, 2(4), 23–31. Duin, A. H., & Pedersen, I. (2021). Writing futures: Collaborative, algorithmic, autonomous. Springer. Edwards-Groves, C. J. (2011). The multimodal writing process: Changing practices in contemporary classrooms. Language and Education, 25(1), 49–64. Ennis, R. H. (2018). Critical thinking across the curriculum: A vision. Topoi, 37, 165-184. Fleckenstein, J., et al. (2024). Do teachers spot AI? Evaluating the detectability of AI-generated texts among student essays. Computers and Education: Artificial Intelligence, 6, 100209. Guilford, J. P. (1967). The nature of human intelligence. McGraw-Hill. Hayes, J. R., & Flower, L. (1980). Identifying the organization of writing processes. In L. Gregg & E. Steinberg (Eds.), Cognitive processes in writing. Erlbaum. Kress, G. (2003). Literacy in the New Media Age. Routledge. Lin, C. H., et al. (2025). Integrating generative AI into digital multimodal composition: A study of multicultural second-language classrooms. Computers and Composition, 75, 102895. Liu, M., Zhang, L. J., & Biebricher, C. (2024). Investigating students’ cognitive processes in generative AI-assisted digital multimodal composing and traditional writing. Computers & Education, 211, 104977. Su, J., & Yang, W. (2023). Unlocking the Power of ChatGPT: A Framework for Applying Generative AI in Education. ECNU Review of Education, 6(3), 355-366. Trinchero, R. (2022). Metodo, atteggiamento, consapevolezza. In Percorsi di ricerca didattica e docimologica: studi in onore di Cristina Coggi (pp. 17–41). FrancoAngeli.