Pretty, Polished, and Wrong

This NotebookLM Deep Dive explores a surprisingly common problem in the age of AI-generated visuals: infographics that look beautiful, professional, and classroom-ready, but quietly teach the wrong thing. Using Theodosis Karageorgakis’ Educraft article and video as a jumping-off point, the hosts unpack why polished layouts, clean typography, and balanced colors can create false confidence, especially when the underlying structure of the information is inaccurate. The episode then walks through a practical two-prompt method for fixing the problem: first forcing the AI to build a rigorous educational analysis in text, then using that factual scaffold to generate a simpler, clearer, cognitively sound visual. Along the way, the conversation connects AI hallucination, visual hierarchy, cognitive load, and multimedia learning theory into a useful workflow for teachers, students, presenters, and anyone who wants AI graphics that do more than look good. They need to actually teach.