Many educators utilize AI-tools to generate classroom materials quickly and efficiently. AI-created readings, worksheets, and listening tasks can provide level-appropriate input rich in target grammar and vocabulary, particularly in ESL contexts where teachers often adapt informational texts for learners. However, the increasing use of such materials raises an important question: how trustworthy is AI-generated educational content - and has the quality changed with recent updates? Research shows exposure to inaccurate information can lead learners to internalize false knowledge through processes such as source confusion and the illusory truth effect. If AI-generated materials contain subtle errors, students may unintentionally learn misinformation - particularly problematic for medical and science students. This study revisits and extends earlier work on the value of AI-generated classroom materials. Subject-matter experts, including doctors and professors, first identified topics within their disciplines, after which reading texts were generated by multiple AI models and evaluated. The consultants assessed the materials for factual accuracy, pedagogical suitability, and potential risk for use as simplified informational texts for language learners. This follow-up study offers insights into when AI-generated materials support learning - and when they undermine it. The findings aim to help educators use AI tools more critically when developing instructional materials.