Spot the Spin — read the facts, not the framing

An x-ray for spin: paste any headline or news link and see exactly how it's manipulating you — scored, explained, and rewritten straight.

Comment

Inspiration

Product people live inside content — industry "insights," competitor PR, founder threads, launch posts. A lot of it isn't written to inform you; it's written to move you. With AI making persuasive copy infinite and free, the distance between "a fact" and "a frame" is collapsing. I wanted an x-ray: paste anything you're reading and see exactly how it's spun — so you decide on the facts, not the framing.

What it does

Spot the Spin is a media-manipulation detector. Paste a headline, an article, or just a news link, and it returns:

  • a Manipulation Index from 0–100,
  • a radar of 8 persuasion tactics (Fear appeal, Outrage bait, Curiosity gap, Exaggeration, Bandwagon, Urgency/scarcity, Guilt-tripping, Punctuation spam),
  • a tactic-by-tactic breakdown with the exact quote that triggered each one and why it's manipulative,
  • a de-spun objective rewrite, and
  • a downloadable "Manipulation Facts" label — like a nutrition label, but for spin.

It's bidirectional: Add Spin turns a plain headline into clickbait so you can watch the playbook being built, and Strip to Facts runs it in reverse.

Who it's for

Builders and product people who need to read industry news, competitor puff pieces, and PR copy on the facts — and who want to check whether their own launch copy is spinning. Plus anyone who wants to read the news with their eyes open.

How I built it

  • Frontend: hand-rolled HTML/CSS/JS with an editorial "fact-check desk" design, a Chart.js radar, and html2canvas to export the shareable label.
  • Backend: Node + Express + TypeScript — static site plus three endpoints (mask / unmask / analyze).
  • AI: Qwen via DashScope's OpenAI-compatible API, with a fixed 8-tactic taxonomy and tolerant JSON parsing so outputs stay consistent across very different inputs.
  • Analytics: Novus (Pendo), auto-loaded from an env var; the three core actions are instrumented as track events.
  • Tooling: built with Claude Code.

Challenges I ran into

  • Fair scoring. A headline in isolation is easy to over- or under-rate. Letting users paste a URL so the model reads the actual article made verdicts noticeably fairer.
  • Consistency. Free-form tactic names broke the radar. A fixed, enumerated 8-tactic taxonomy plus alias-matching is what makes the radar and the label trustworthy.
  • Graceful AI. Third-party endpoints drift in format; tolerant JSON extraction keeps the UI solid.

Accomplishments I'm proud of

A live tool a stranger can use in one click, a memorable shareable artifact (the Manipulation Facts label), and a bidirectional model that teaches the playbook instead of just flagging it.

What I learned

Turning an abstract "manipulation score" into a concrete, shareable object changed how people react to it. And grounding analysis in the real article — not just the headline — matters more than any prompt tweak.

What's next for Spot the Spin

A browser extension that scores headlines in your feed, a "compare two sources on the same story" view, and an "Honest Copy" mode so builders can check their own launch copy before they ship.

Built With

Share this project:

Updates