This paper presents findings from a year-long project in which Grade 4 pupils engaged with AI chat assistants and image generators in a Catholic religious learning environment (RLE). Using the GDPR-compliant platform Fobizz and school-issued iPads, learners explored the narratives of Jacob and Esau (Gen 25-36) and Moses at the burning bush through dialogic AI interaction, targeted prompt development, and AI-generated visualisations documented in digital learning journals. The project addresses how AI tools can enable inclusive access to religious knowledge—particularly for children from less media-literate households—by embedding personalised devices within a media education concept grounded in Tulodziecki's action-theoretical model of media competence. Theologically, the unit employs the Bibliolog distinction between "black fire" (scriptural text) and "white fire" (interpretive space) as a framework for evaluating AI-generated content: pupils differentiated textual evidence from machine-produced interpretation and assessed outputs for theological adequacy, fidelity, and potential distortions. Structured peer feedback fostered metacognitive reflection and critical thinking as key competencies for navigating generative AI in religious education (cf. Chrostowski 2024). Results show that AI-supported sequences were perceived as particularly memorable, that prompt precision correlated with output quality, and that pupils developed a differentiated awareness of AI's capacities and limitations within the hermeneutical process. The paper contributes to the panel's inquiry into how AI tools are received in religious education and whether generative AI can foster inclusive, critically reflective access to religious knowledge in primary RLE.