Permitting AI in reading preparation did not produce clear gains
AI-permitted and non-AI reading preparation showed no statistically significant differences in quizzes, examination scores, preparation time or perceived understanding. Students mainly sought vocabulary and sentence-level help.
Why it matters for leaders
For a reading pilot, require an initial interpretation and a short explanation of any revision after assistance. Measure comprehension on a new passage without AI, alongside preparation time.
The evidence
The within-participant study enrolled 83 Japanese university students. Complete paired analyses included 64 quiz records and 75 examination records. Usage logs and reflections showed less frequent checking of students' own interpretations than requests for explanations.
Who was studied
Non-English-major undergraduates at a private Japanese university.
What the evidence cannot tell us
One university, different reading materials, missing records and varied tools limit inference. A non-significant result establishes neither equivalence nor safety. University English-language reading does not directly represent elementary literacy.
Read the original sources
Summary and leadership analysis from the Week ending 2026-09-18 brief. The edition’s selection notes explain source access and evidence limitations.