Reading proficiency among United States eighth graders dropped to 30% in 2024, marking a 30-year low as learners struggle with what educators describe as a "comprehension wall." While foundational literacy investments in phonics, phonemic awareness, and fluency show success through early primary grades, student performance drops sharply in middle school as texts demand deeper vocabulary, background knowledge, and the ability to process complex sentence structures across science, maths, and humanities subjects.

Rather than using generic chatbots, educators and researchers advocate for custom-built literacy tools grounded in the science of reading. Citing findings from the Brookings Institution, which caution that AI tools must "teach, not tell" to be effective, emerging implementations focus on continuous diagnostic support. Instead of relying on isolated standardised assessments, these tools analyze students' day-to-day reading work, pinpoint specific comprehension failures, and offer targeted interventions, such as prompting learners to reread specific passages or delivering context-specific vocabulary support.

This shift signals a broader move in educational technology away from blunt assessment scores toward granular, real-time diagnostic scaffolding in mixed-ability classrooms. For South African educators managing the challenging transition from early-grade decoding to language-intensive Senior Phase curricula, targeted AI scaffolding models demonstrate how digital tools can assist differentiated reading instruction without reducing academic rigor.