Teacher Test Shows AI Detectors Giving Wildly Inconsistent Results on Human Essays
ruthstarkman · x · 2026-08-30
A teacher tested five different AI detection programs on an in-class student essay (approx. 800 words) written without AI assistance. The results were extremely inconsistent, ranging from "100% AI" to "likely 100% human". This experiment highlights the severe unreliability and lack of accuracy in current AI detection tools.
More from Apps
- Opinion: Omarchy is Far From Ready, Linux Desktop Year Not Here Yet — steren · 2026-08-30
- User switches to GLM 5.3 Flash for docs, sees zero quality drop and lower cost than GPT — abacaj · 2026-08-30
- Case Study: Using AI to Plan a $1M ARR Strategy for a Blogger — huangyun_122 · 2026-08-30
- Cardgo: An AI Card Context Manager for Cross-Model Usage — dieutrinhngoc · 2026-08-30
- Hailuo MiniMax Design passes third prototype test: Gaudí-style e-commerce UI — Hailuo_AI · 2026-08-30
- Open-Source GEO Tool Optimizes Websites for ChatGPT and Claude Citations — tom_doerr · 2026-08-30