AI detectors

Compare AI detectors head to head

Every pair of the seven AI detectors we tested, compared on the same 756 texts: catch rate, false positives, AI models and kinds of writing.

• Desklib AI text detector vs Fast-DetectGPT (Qwen2.5-1.5B)
  Flags human text: 6.3% vs 15.9%
• Fakespot AI text detector vs Fast-DetectGPT (Qwen2.5-1.5B)
  Flags human text: 34.9% vs 15.9%
• Desklib AI text detector vs Fakespot AI text detector
  Flags human text: 6.3% vs 34.9%
• Fast-DetectGPT (Qwen2.5-1.5B) vs MAGE
  Flags human text: 15.9% vs 19.0%
• Desklib AI text detector vs MAGE
  Flags human text: 6.3% vs 19.0%
• E5-small LoRA detector vs Fast-DetectGPT (Qwen2.5-1.5B)
  Flags human text: 50.0% vs 15.9%
• Desklib AI text detector vs E5-small LoRA detector
  Flags human text: 6.3% vs 50.0%
• Fakespot AI text detector vs MAGE
  Flags human text: 34.9% vs 19.0%
• Fast-DetectGPT (Qwen2.5-1.5B) vs OpenAI RoBERTa detector
  Flags human text: 15.9% vs 4.8%
• Desklib AI text detector vs OpenAI RoBERTa detector
  Flags human text: 6.3% vs 4.8%
• E5-small LoRA detector vs Fakespot AI text detector
  Flags human text: 50.0% vs 34.9%
• Fast-DetectGPT (Qwen2.5-1.5B) vs TMR (Target Mining RoBERTa)
  Flags human text: 15.9% vs 83.3%
• Desklib AI text detector vs TMR (Target Mining RoBERTa)
  Flags human text: 6.3% vs 83.3%
• E5-small LoRA detector vs MAGE
  Flags human text: 50.0% vs 19.0%
• Fakespot AI text detector vs OpenAI RoBERTa detector
  Flags human text: 34.9% vs 4.8%
• Fakespot AI text detector vs TMR (Target Mining RoBERTa)
  Flags human text: 34.9% vs 83.3%
• MAGE vs OpenAI RoBERTa detector
  Flags human text: 19.0% vs 4.8%
• E5-small LoRA detector vs OpenAI RoBERTa detector
  Flags human text: 50.0% vs 4.8%
• MAGE vs TMR (Target Mining RoBERTa)
  Flags human text: 19.0% vs 83.3%
• E5-small LoRA detector vs TMR (Target Mining RoBERTa)
  Flags human text: 50.0% vs 83.3%
• OpenAI RoBERTa detector vs TMR (Target Mining RoBERTa)
  Flags human text: 4.8% vs 83.3%