The contemporary era is marked by a strong acceleration in knowledge production processes and an unprecedented proliferation of information. In this context, the diffusion of generative artificial intelligence (AI) highlights the need for pedagogical and critical work that does not focus exclusively on the technology itself, but rather on its educational and cultural implications.

In particular, teachers and future teachers are increasingly required to address these transformations within competence-based educational design. A crucial aspect concerns the growing role of content—both textual and visual—often produced or mediated by AI. While such content is highly persuasive, it frequently conveys partial, fragmented, or simplified representations of complex phenomena. As a result, knowledge construction increasingly relies on visual and audiovisual languages.

Research by Chang, Yang, and Wong (2024) shows that AI has become an integral part of everyday life and that, already in primary school, children develop preconceptions and representations related to AI. In this scenario (Manovich, 2016; Panciroli & Rivoltella, 2023) the role of algorithms in shaping information flows. Social media environments, where images and short videos dominate both the production and circulation of knowledge, represent a paradigmatic example of this process.

Although the human brain is intrinsically visual (Robertson, 2011), formal schooling has historically privileged verbal language, marginalizing systematic education in visual creativity and visual language. This imbalance becomes particularly problematic in contexts where knowledge is increasingly mediated through images and videos. As Lakoff and Johnson (1980) and Black (1983) argue, a fundamental characteristic of human thinking is its metaphorical and creative nature: meaning is constructed through processes of projection, analogy, and imagination. Images, therefore, are not merely illustrative supports, but cognitive devices that actively participate in shaping plural, situated, and culturally embedded representations of the world.

Within this framework, AI emerges as a central actor in meaning-making processes. The systematic review by Heigl (2025) highlights how professional artists recognize AI as a powerful tool for creativity, while simultaneously stressing the importance of human agency in the construction of meaning. Similarly, Yaşar et al. (2025) show that university students tend to conceptualize creativity not only as an individual capacity to generate novelty, but as a competence situated within hybrid cultural ecosystems in which AI acts as a mediator. These authors frame prompt literacy not as a purely technical skill, but as an educational, cultural, and ethical issue. From a more instrumental perspective, Mujica-Sequera (2025) defines prompting as a lever of cognitive efficiency aimed at formulating effective queries, optimizing data processing, and identifying patterns.

AI currently functions as a synthesizer of information: the images it generates tend to be generic, idealized, and stereotyped, privileging iconic and recognizable elements while excluding the ordinary, the marginal, and the everyday (Manovich, 2025). Converging evidence is provided by Freixa et al. (2025), whose analysis of 600 images shows that AI-generated visuals respond more precisely and coherently to textual prompts. However, this very alignment reinforces dominant visual schemas, making stereotypes more efficient, reproducible, and standardized, while reducing the semantic and narrative complexity of representations.

In this scenario the DigComp 3.0 (Cosgrove & Cachia, 2025) highlights a strong emphasis on critical digital literacy, where the production and interpretation of texts and images require reflective and critical thinking in order to counter processes of semantic homogenization. In light of this theoretical framework, the present paper analyses video stories created by pre-service primary teachers enrolled in the Primary Teacher Education programme at the University of Turin, within the course Educational Technologies. The study is situated in an innovative learning environment that integrates artificial intelligence both as a tool for digital artefact production and as an object of critical reflection.

The study adopts a qualitative research design, supported by descriptive quantitative analysis, and is grounded in a deliberately designed innovative learning environment. The research sample consists of 170 educational video stories produced by second-year students enrolled in the Educational Technologies course during the 2024/2025 academic year. The activity was embedded in regular course assessment, ensuring the inclusion of the entire cohort and avoiding self-selection bias. The learning activity was structured through a shared pedagogical design framework. Students were provided with a design grid guiding the development of each video story, requiring them to define the title, narrative incipit, educational objectives, target audience, didactic function, visual and narrative coherence, and the AI prompts used for image generation. To ensure both creative freedom and analytical comparability, students selected one narrative incipit from a curated set of ten prompts designed to encourage imaginative and metaphorical storytelling. The choice of storytelling as the central pedagogical approach was intentional. Storytelling was adopted as it fosters reflective and critical engagement with the content to be communicated, requiring students to make explicit decisions about meaning, perspective, and relevance.t (Bruner, 1996). A key pedagogical requirement was the explicit request to create images that went beyond simple referential representation. Students were encouraged to generate metaphorical and evocative images capable of conveying abstract concepts, emotions, and educational meanings through visual symbolism. The analytical process was articulated across two main dimensions: visual analysis and narrative-textual analysis. The visual analysis followed a semiotic approach and was conducted using an ad hoc image analysis grid specifically developed for this study. This grid enabled the systematic classification of AI-generated images according to: (a) type of representation (metaphorical, referential, decorative); (b) type of meaning conveyed (symbolic, evocative, illustrative); (c) recurrence of visual representations, identifying dominant iconographic elements The narrative and textual analysis was conducted using qualitative content analysis. Video narratives were examined through the lens of narrative grammar, focusing on narrative sequence, character roles, plot development, and thematic organisation. Particular attention was paid to the type of incipit selected and its influence on narrative progression. Emerging themes were identified through inductive coding, allowing for the identification of recurring narrative structures and educational motifs. By combining a semiotic image analysis grid with qualitative content analysis of narratives, the methodology enabled a systematic exploration of how generative AI mediates visual meaning-making and narrative construction within an innovative learning environment.

The analysis is currently ongoing; however, findings from a first cycle of qualitative content analysis focusing on textual prompts and narrative structures already reveal significant patterns in AI-supported storytelling. Based on the preliminary analysis of 170 video artefacts, students’ textual prompts show a strong tendency towards structural regularity. In particular, the majority of prompts follow a recurrent syntactic sequence (e.g. subject–action–setting–style), with limited variation in linguistic formulation and descriptive depth. This suggests an early convergence towards perceived “effective” or “safe” prompting strategies when interacting with generative AI systems. From a narrative perspective, initial content analysis indicates that over two thirds of the video stories adopt linear and familiar narrative trajectories, characterised by emotionally stable characters, predictable developments, and limited narrative tension. The choice of incipit plays a relevant role in this convergence: a small subset of incipit types was selected by a majority of students, and these incipits often oriented the narratives towards reassuring, emotionally positive, and educationally conventional storylines. Emerging themes frequently revolve around exploration, discovery, care, and personal growth, with relatively limited narrative divergence. These preliminary findings provide a partial but substantial response to the first research questions. They shed light on how students use generative AI through textual prompts and narrative construction within an innovative learning environment, and they highlight recurring narrative and linguistic patterns that point towards processes of standardisation. The analysis of AI-generated images is currently ongoing and will complement these findings by offering a systematic semiotic examination of visual representations and imaginaries.

Black, M. (1983). Modelli, archetipi, metafore. Pratiche. Bruner, J. (1996). The culture of education. Harvard University Press. Chang, X., Yan, Z., & Wong, G. K. (2024). Young children’s understanding of artificial intelligence: A draw-a-picture analysis. British Journal of Educational Technology. Cosgrove, J., & Cachia, R. (2025). DigComp 3.0: European digital competence framework. European Commission, Joint Research Centre. Freixa, P., Redondo-Arolas, M., Codina, L., & Lopezosa, C. (2025). AI, stock photography, and image banks: Gender biases and stereotypes. Hipertext.net, (30). Heigl, R. (2025). Generative artificial intelligence in creative contexts: A systematic review and future research agenda. Management Review Quarterly, 1–38. Lakoff, G., & Johnson, M. (1980). Metaphors we live by. University of Chicago Press. (ed. it.: Lakoff, G., Johnson, M., & Violi, P. (1981). Metafora e vita quotidiana. Bompiani). Manovich, L. (2016). Software takes command. Bloomsbury Academic. Manovich, L. (2025). Unreliable past: Constructing memory with generative AI. Leonardo, 58(5), 477–483. Mujica-Sequera, R. M. (2025). AI prompts: Tools for optimizing scientific research. Revista Tecnológica-Educativa Docentes 2.0, 18(1), 267–277. Panciroli, C., & Rivoltella, P. C. (2023). Pedagogia algoritmica: Per una riflessione educativa sull’intelligenza artificiale. Scholé. Robertson, I. (2011). The mind’s eye. Random House. Yaşar, İ., Arslan Selçuk, S. E. M. R. A., & Alaçam, S. (2025). Use of artificial intelligence and prompt literacy in architectural education. New Design Ideas, 9(1), 248–268.