Mathematical anxiety affects approximately 20-30% of primary school students, negatively influencing academic achievement and fostering persistent avoidance of mathematical activities. This study presents a systematic comparative analysis of three generative AI platforms (Gemini Story Book, Grok, and NotebookLM) used to create illustrated mathematical fairy tales as an innovative preventive approach for children aged 6-10. Twelve original mathematical fairy tales were created on each platform (total sample n=36 narratives), covering key topics of the primary mathematics curriculum. Two independent experts with experience in primary mathematics education and educational technologies conducted a systematic evaluation of the generated content according to a developed rubric with four criteria: mathematical accuracy, pedagogical relevance, narrative coherence, and visual quality (scale of 1-3 points). A total of 36 original narratives (n = 36), with twelve tales generated per platform, addressed key topics of the primary mathematics curriculum. The results demonstrate statistically significant differences between platforms: Gemini Story Book achieved the highest scores in all categories (mathematical accuracy M=2.71, SD=0.464; pedagogical relevance M=2.79, SD=0.415), while Grok revealed critical issues with the accuracy of automatic narration (M=1.42, SD=0.504), and Notebook LM demonstrated a systematic drift in mathematical content (M=2.17, SD=0.565). Based on empirical findings, a five-step scientifically grounded workflow was developed that combines AI efficiency with pedagogical quality control. The results are relevant for the development of educational technologies, the prevention of math anxiety, and the integration of generative artificial intelligence into primary education.