Enhancing the Reading Comprehension of Grade 5 Learners Through AI-Driven Levelled Reading Instruction

Abstract

This action research addressed a critical literacy gap identified during the June 2026 Phil-IRI diagnostic assessment at Pangapisan Elementary School, where more than sixty percent of Grade 5 learners performed at the Frustration Level. Because static, uniform reading materials could not accommodate the highly varied learning paces of these learners, and because a single teacher could not feasibly plan and deliver nineteen distinct, individualized reading pathways at once, the study evaluated an AI-driven levelled reading instruction intervention using a pre-experimental, one-group pretest-posttest design. The nineteen participants formed a homogeneous group of Grade 5 learners from Section D identified as non-readers who struggled specifically with Consonant-Vowel-Consonant word patterns. Treating the entire cohort as a single experimental unit allowed the group to serve as its own control, avoiding the ethical problem of withholding remedial instruction from vulnerable public school learners in order to maintain a separate untreated comparison group. Data were gathered using a thirty-item parallel-form comprehension test adapted from Phil-IRI standards and were analyzed using descriptive and inferential statistics at α = 0.05. Results showed a significant upward shift in comprehension scores, with learners moving from the Frustration Level toward the Instructional Level and, for a number of learners, the Independent Level, following personalized text calibration. Beyond the numerical gains, the intervention was associated with reduced reading anxiety arising from self-paced progression, greater learner ownership of reading milestones, and more efficient use of instructional time, since the AI platform assumed responsibility for routine text levelling while the teacher concentrated on higher-level remediation.