We tested GLM 5.2 on AEO writing tasks and measured a 31% hallucination rate. Get our step-by-step methodology, results, and what it means for your AI content