The AI Information Trust Benchmark

Enter a prompt on any societal or policy topic. We'll run it through AI models and evaluate each response for bias and information quality.

0 / 5000
Flagship
Lite
Anthropic
OpenAI
Google
xAI
DeepSeek

How it works

  1. Enter a prompt on a political or controversial topic
  2. Select 1-2 AI models to test
  3. Each response is automatically graded on bias and quality criteria
  4. Review the results and share your own assessment

What we measure

Bias

  • Tone & Framing — Is the language neutral?
  • Balance — Are perspectives represented fairly?

Quality

  • Factual Accuracy — Are claims correct and sourced?
  • Substantive Engagement — Does it actually address the topic?
  • Confidence Calibration — Does certainty match the evidence?

View evaluation methodology →