ResearchThe Decoder· Jun 25, 2026
Authors Guild Tests Show Mixed Results for AI Detectors on Human Text
The Authors Guild evaluated five AI detectors using human-written texts. Pangram and Grammarly correctly classified every sample as human-written. Sidekicker and ZeroGPT incorrectly labeled the same texts as AI-generated. The group notes that professional writing can resemble AI output because models were trained on similar material.
Key points
- →Pangram and Grammarly identified all human texts correctly.
- →Sidekicker and ZeroGPT flagged human texts as AI-generated.
- →Professional writing may statistically resemble AI output due to training data.
Mentioned
PangramGrammarlySidekickerZeroGPT
Related stories
Google Losing Top AI Researchers to Rival CompaniesThe Decoder · Research→AI Researchers Depart Google for Anthropic and RivalsTechCrunch · Research→Research Shows Reasoning Improves Knowledge Recall in Large Language ModelsGoogle Research · Research→Pangram CEO Says Language Models Produce Clustered ArgumentsThe Decoder · Models→Hugging Face Launches FFASR Leaderboard for Real-World ASR BenchmarkingHugging Face · Research→AI Loops Extend Agentic Systems with Continuous Agent SwarmsTechCrunch · Research→
This is an original summary by Dhanasvi's agents based on The Decoder's public feed. For the complete article, visit the original source. Trademarks and article copyright belong to their owners.