Learning print-speech sound correspondences is a crucial step at the beginning of reading acquisition and often impaired in children with developmental dyslexia. Despite increasing insight into audiovisual language processing, it remains largely unclear how integration of print and speech develops at the neural level during initial learning in the first years of schooling. To investigate this development, 32 healthy, German-speaking children at varying risk for developmental dyslexia (17 typical readers and 15 poor readers) participated in a longitudinal study including behavioral and fMRI measurements in first (T1) and second (T2) grade. We used an implicit audiovisual (AV) non-word target detection task aimed at characterizing differential activation to congruent (AVc) and incongruent (AVi) audiovisual non-word pairs. While children's brain activation did not differ between AVc and AVi pairs in first grade, an incongruency effect (AVi > AVc) emerged in bilateral inferior temporal and superior frontal gyri in second grade. Of note, pseudoword reading performance improvements with time were associated with the development of the congruency effect (AVc > AVi) in the left posterior superior temporal gyrus (STG) from first to second grade. Finally, functional connectivity analyses indicated divergent development and reading expertise dependent coupling from the left occipito-temporal and superior temporal cortex to regions of the default mode (precuneus) and fronto-temporal language networks. Our results suggest that audiovisual integration areas as well as their functional coupling to other language areas and areas of the default mode network show a different development in poor vs. typical readers at varying familial risk for dyslexia.