Green-energy literacy is often weak in multilingual primary classrooms, where pupils struggle to distinguish renewable from non-renewable systems and to master technical terms. Few classroom trials jointly examine bilingual augmented reality (AR) and the cognitive-load mechanisms that may explain learning gains. This study tested whether Arabic–French AR flashcards enhance Grade-6 pupils’ recognition of renewable-energy systems, optimise cognitive-load profiles, and increase curiosity about related careers. We ran a cluster-randomised, post-test-only trial with 116 pupils from two public schools in Casablanca. Experimental classes used five AR flashcards (solar photovoltaic, wind turbine, micro-hydro, biogas digester, natural-gas plant) with dual-language labels and brief narration; comparison classes used a content-matched bilingual worksheet. Outcomes were a 12-item recognition quiz, a 10-item Cognitive Load Questionnaire (intrinsic, extraneous, germane), a 6-item career-interest scale, and a neighbourhood “photo-hunt” transfer task. ANCOVA showed a + 23.4 pp advantage for AR on recognition (p < 0.001; η2 = 0.24; d = 0.93). AR learners reported lower extraneous load (− 0.58 SD, p < 0.01) and higher germane load (+ 0.37 SD, p < 0.05), with no difference in intrinsic load; career interest increased (+ 0.42 SD, p < 0.05). The AR group submitted 2× as many photo-hunt entries and achieved higher caption accuracy (72% vs 34%; χ2(1, N = 93) = 10.37, p < 0.01. Low-cost bilingual AR can deepen knowledge, improve cognitive efficiency, and extend learning beyond class. Future work should examine retention and objective load measures to isolate active design ingredients.