Slant Detective

Tracking since Apr 23, 2026.

Changelog

• May 5, 2026
  description

  See how any news article is slanted — right in the article itself.
  
  Click the toolbar icon on any news page and a side panel opens with a structured readout: the overall lean, the loaded words, and the specific phrases doing the persuading. Every flagged phrase is underlined in the article so you can see the evidence in context.
  
  No account. No server. Your article never leaves your browser.
  
  ────────────────────────
  
  FREE — instant, no setup, no key required
  An entirely in-browser analysis:
  • A 2,100-word bias lexicon from peer-reviewed research (BABE)
  • Attribution-verb skew ("said" vs. "alleged" vs. "admitted")
  • Headline-vs-body drift
  • Hedge-word count
  • An outlet-level lean label (e.g. "Lean Left," "Center")
  
  IN-DEPTH ANALYSIS — paste your own API key
  Paste a key from Anthropic, OpenAI, or Google and unlock:
  • An overall intensity score and direction
  • A four-dimension breakdown (word choice, framing, headline, source mix)
  • Evidence phrases with a one-sentence reason for each
  • Dotted underlines on the actual biased phrases in the article
  • Hover for the reason, click to scroll to the phrase
  
  Your key stays in your browser. The call goes straight from your browser to your provider. We never see the article or the key.
  
  ────────────────────────
  
  COST
  Free tier is free forever. In-depth analysis uses your own API key, and you choose the provider: Gemini 2.5 Flash runs about $0.0002 per article — well under $1/month for a typical reader. Claude Haiku runs about $0.0018 per article — roughly $1–3/month. OpenAI GPT-5-mini sits in between. You pay your provider directly — we take no cut.
  
  OPEN SOURCE
  Full source, the bias-word lexicon, and the outlet-label dataset are all AGPL-3.0 on GitHub. You can verify exactly what data leaves your device and where it goes. We run no backend. No account database. No advertising.
  
  ────────────────────────
  
  WHAT'S NEW IN v1.1
  • Choose your provider: Anthropic, OpenAI, or Google Gemini
  • Percentile context — see how the article's score compares to other news domains
  • Dotted underlines on biased phrases, right in the article
  • Hover tooltips with category and reason
  • Works on AMP pages and JS-hydrated sites (NYT, Breitbart, and more)
  • Info-icon explanations on every dimension label
  • Plain-English rationale on the signal summary

  Click the toolbar icon on any news page and a side panel opens with a structured readout: the overall lean, the loaded words, and the specific phrases doing the persuading. Every flagged phrase is underlined in the article so you can see the evidence in context.
  
  No account. No server. Your article never leaves your browser.
  
  ────────────────────────
  
  FREE — instant, no setup, no key required
  An entirely in-browser analysis:
  • A 2,100-word bias lexicon from peer-reviewed research (BABE)
  • Attribution-verb skew ("said" vs. "alleged" vs. "admitted")
  • Headline-vs-body drift
  • Hedge-word count
  • An outlet-level lean label (e.g. "Lean Left," "Center")
  
  IN-DEPTH ANALYSIS — paste your own API key
  Paste a key from Anthropic, OpenAI, or Google and unlock:
  • An overall intensity score and direction
  • A four-dimension breakdown (word choice, framing, headline, source mix)
  • Evidence phrases with a one-sentence reason for each
  • Dotted underlines on the actual biased phrases in the article
  • Hover for the reason, click to scroll to the phrase
  
  Your key stays in your browser. The call goes straight from your browser to your provider. We never see the article or the key.
  
  ────────────────────────
  
  COST
  Free tier is free forever. In-depth analysis uses your own API key, and you choose the provider: Gemini 2.5 Flash runs about $0.0002 per article — well under $1/month for a typical reader. Claude Haiku runs about $0.0018 per article — roughly $1–3/month. OpenAI GPT-5-mini sits in between. You pay your provider directly — we take no cut.
  
  OPEN SOURCE
  Full source, the bias-word lexicon, and the outlet-label dataset are all AGPL-3.0 on GitHub. You can verify exactly what data leaves your device and where it goes. We run no backend. No account database. No advertising.
  
  ────────────────────────
  
  WHAT'S NEW IN v1.1
  • Choose your provider: Anthropic, OpenAI, or Google Gemini
  • Percentile context — see how the article's score compares to other news domains
  • Dotted underlines on biased phrases, right in the article
  • Hover tooltips with category and reason
  • Works on AMP pages and JS-hydrated sites (NYT, Breitbart, and more)
  • Info-icon explanations on every dimension label
  • Plain-English rationale on the signal summary

• Apr 27, 2026
  description

  Slant Detective analyzes any news article you're reading for language bias — right in the page you're already on. Click the toolbar icon on any article and a side panel opens instantly with results. No sign-up. No account. No server that ever sees your reading.
  
  **Layer 1 — works immediately, no setup required.** Even without an API key, Slant Detective runs entirely in your browser: it scans the article body against a research-backed bias-word lexicon (based on the BABE dataset, 2,100+ loaded terms), tallies attribution-verb skew (how differently sources are described — "said" vs. "alleged" vs. "admitted"), measures headline drift from the article body, and counts hedge words. The source label (e.g., "Lean Left," "Center") comes from our own open-source dataset, AGPL-3.0, bundled in the extension.
  
  **Layer 2 — full rubric with inline highlights.** Paste your own Anthropic API key (takes 3 minutes — we walk you through it) and you unlock the full analysis: an overall intensity score and political-direction tilt, a four-dimension breakdown (word choice, framing, headline slant, source mix), a list of specific evidence phrases with one-sentence explanations, and dotted underlines on the actual biased phrases in the article. Hover any underline for the reason. Click any evidence row in the panel and the page scrolls to the phrase. Your key stays on your device; the call goes straight from your browser to Anthropic — we never see the article or your key.
  
  **Cost to you:** Layer 1 is completely free. Layer 2 uses your own Anthropic key at roughly $0.006 per article (Claude Haiku 3.5, ~3,000 input tokens + ~800 output tokens). A typical reader spends under $2/month.

  See how any news article is slanted — right in the article itself.
  
  Click the toolbar icon on any news page and a side panel opens with a structured readout: the overall lean, the loaded words, and the specific phrases doing the persuading. Every flagged phrase is underlined in the article so you can see the evidence in context.
  
  No account. No server. Your article never leaves your browser.
  
  ────────────────────────
  
  FREE — instant, no setup, no key required
  An entirely in-browser analysis:
  • A 2,100-word bias lexicon from peer-reviewed research (BABE)
  • Attribution-verb skew ("said" vs. "alleged" vs. "admitted")
  • Headline-vs-body drift
  • Hedge-word count
  • An outlet-level lean label (e.g. "Lean Left," "Center")
  
  IN-DEPTH ANALYSIS — paste your own API key
  Paste a key from Anthropic, OpenAI, or Google and unlock:
  • An overall intensity score and direction
  • A four-dimension breakdown (word choice, framing, headline, source mix)
  • Evidence phrases with a one-sentence reason for each
  • Dotted underlines on the actual biased phrases in the article
  • Hover for the reason, click to scroll to the phrase
  
  Your key stays in your browser. The call goes straight from your browser to your provider. We never see the article or the key.
  
  ────────────────────────
  
  COST
  Free tier is free forever. In-depth analysis uses your own API key, and you choose the provider: Gemini 2.5 Flash runs about $0.0002 per article — well under $1/month for a typical reader. Claude Haiku runs about $0.0018 per article — roughly $1–3/month. OpenAI GPT-5-mini sits in between. You pay your provider directly — we take no cut.
  
  OPEN SOURCE
  Full source, the bias-word lexicon, and the outlet-label dataset are all AGPL-3.0 on GitHub. You can verify exactly what data leaves your device and where it goes. We run no backend. No account database. No advertising.
  
  ────────────────────────
  
  WHAT'S NEW IN v1.1
  • Choose your provider: Anthropic, OpenAI, or Google Gemini
  • Percentile context — see how the article's score compares to other news domains
  • Dotted underlines on biased phrases, right in the article
  • Hover tooltips with category and reason
  • Works on AMP pages and JS-hydrated sites (NYT, Breitbart, and more)
  • Info-icon explanations on every dimension label
  • Plain-English rationale on the signal summary

• Apr 27, 2026
  short_description

  Per-article media-bias analysis. Open source, AGPL-3.0. No account, no article data stored on our servers.

  See how any news article is slanted — the lean, the loaded words, the phrases doing the work. In the article itself.

• Apr 27, 2026
  host_permissions

  https://api.anthropic.com/*, https://sd-telemetry.rabbit-factory.workers.dev/*

  https://api.anthropic.com/*, https://api.openai.com/*, https://generativelanguage.googleapis.com/*, https://sd-telemetry.rabbit-factory.workers.dev/*

Permissions & access

Permissions

sidePaneltabsactiveTabstoragescriptingalarms

Host access

https://api.anthropic.com/*, https://api.openai.com/*, https://generativelanguage.googleapis.com/*, https://sd-telemetry.rabbit-factory.workers.dev/*

Screenshots