See the bias in your news.

Your AI watchdog for objectivity.

How It Works — Live Example

Claude: The headline uses the loaded verb 'dragging' (implying coercion/unwilling participation) and frames a complex geopolitical situation as a yes/no question, presenting an interpretation rather than a verifiable fact about whether paramilitaries are actually 'dragging' Iraq into conflict. | OpenAI: The headline uses the phrase 'dragging Iraq into the war,' which implies a negative action and evokes an emotional response, while also posing a question that suggests speculation about the motives of Iran-allied paramilitaries.

🟣Claude
Loaded Language65
Framing55
Factuality40
🟢OpenAI
Loaded Language80
Framing60
Factuality70
LLM Agreement: 83%Final score is the average of both AI engines

How scores work: Each article is scored 0–100. Headline language, framing, and factuality account for 75% of the score; lead image analysis (emotional manipulation, sensationalism) accounts for 25%. On topic pages, each channel also shows its directional lean — the perspective its headlines most frequently favor (e.g. “pro-Palestinian”, “pro-government”). A ⚠ Coverage Bias alert appears when a channel both over-covers a topic (2×+ the peer average) and scores below 40/100. Learn more →

Disclaimer: All scores and labels are generated by AI and may be inaccurate. Analysis quality improves over time as more data is collected and models are refined. Scores should be used as a general guide, not as definitive judgments of journalistic quality.