About Spin Detector
Same story. Multiple outlets. Measurable bias.
What Is Spin Detector?
Spin Detector is an automated media bias tracker that analyzes how 55 major news outlets cover the same political stories. Every day, our pipeline fetches hundreds of headlines, clusters them by topic, and uses AI to score each headline's political language on a −5 to +5 scale — from Far Left to Far Right, with 0 representing a neutral center.
The goal is simple: give readers a concrete, data-driven way to see how language shapes political narratives across the media spectrum.
How the Bias Score Works
Each headline is scored from −5 to +5 based on linguistic signals detected by AI analysis. Zero represents neutral, centrist language — negative scores lean left, positive scores lean right:
Heavy progressive framing, emotionally charged language
Progressive framing, sympathetic to left causes
Neutral verbs, balanced sourcing, minimal ideological signals
Conservative framing, sympathetic to right causes
Heavy conservative framing, charged language
What Signals We Detect
- Word choice — "undocumented" vs "illegal alien"; "gun safety" vs "gun grab"
- Verb loading — "enforcement" (neutral) vs "raids" (left-charged) vs "crackdown" (right)
- Victim framing — who is portrayed as harmed vs threatening
- Source trust — implicit alignment with certain institutions or advocacy groups
- Emphasis — what information leads a headline vs what is omitted
- Qualifiers — "controversial" applied selectively to certain policies
Claude vs Grok — Model Comparison
Bias scoring is performed independently by two AI models: Claude (Anthropic) and Grok (xAI). Both models receive identical prompts and the same headlines, but score them without seeing each other's results. The Model Wars tab shows where they agree and where they diverge — a useful check on whether the bias signal is robust or model-dependent.
Where Claude and Grok reach similar scores independently, confidence in the rating is higher. Wide divergence suggests the framing is ambiguous or that the two models weight different linguistic signals.
The Outlets We Track
We currently track 54 major English-language news outlets across the political spectrum, from HuffPost and Vox on the left to Breitbart and The Federalist on the right, with the Associated Press and Reuters providing a neutral wire-service baseline in the center. Broadcast networks (ABC, NBC, CBS) and commentary outlets (The Atlantic, Reason) round out the US picture. Financial coverage spans Bloomberg, Yahoo Finance, MarketWatch, Forbes, CNBC, Business Insider, the Financial Times, and the Wall Street Journal. US regional papers cover major metros — LA Times, Chicago Tribune, Boston Globe, Houston Chronicle, Miami Herald, Tampa Bay Times, Star Tribune, and Charlotte Observer. International English-language outlets include BBC, Al Jazeera, The Guardian, The Independent, Sky News, Daily Mail, Metro, The Telegraph, The Times (London), Financial Times, CBC News (Canada), and Times of Israel.
Headlines are fetched daily via NewsAPI and RSS feeds. Each outlet has an expected bias range based on established media research — scores significantly outside that range on a given day are flagged as notable.
Important Limitations
Bias scores are generated by AI language model analysis and represent computational estimates of linguistic patterns only.
- Scores do not constitute editorial opinion or factual ratings
- Headline-level analysis may not reflect a full article's content
- Individual article scores may reflect topic-specific framing, not a publication's overall stance
- AI models can have their own biases that affect scoring
- This tool is designed for educational purposes only
Contact
Questions, feedback, or press inquiries: piers@spindetector.com