See the loaded words

Bias Mirror

Hold a headline up to the light and see the loaded words. Bias Mirror highlights charged language, scores its neutrality, and suggests a plainer rewrite — training the eye to spot spin anywhere.

Runs on your device — never sent to a server.

Or pick a sample

Under the light

Officials SLAM reckless plan that could DEVASTATE families

Neutrality

?
Every headline starts at 100. Each loaded word, ALL-CAPS shout, and pile-on of exclamation marks subtracts points (heavier framing costs more). The result is a transparent 0–100 reading of how neutral the wording is — not a judgment of whether the claim is true. 70+ reads neutral, 40–69 mixed, under 40 loaded.

28/100

loaded
?
The verdict is a plain-language band for the score: neutral (70+), mixed (40–69), or loaded (under 40). It summarizes how much the wording leans on emotion rather than information.

What kind of framing

?
Loaded language comes in flavors. Bias Mirror sorts each tell into a category so you can see the technique: charged words, doom/hype, intensifiers (heat, not info), absolutist sweeps (“everyone”, “never”), and ALL-CAPS shouting.

Charged words: 3ALL-CAPS shouting: 2

Loaded words (5)

?
Each row is one rhetorical “tell” we matched, with the reason it was flagged. These are wording patterns, never political positions — the same rules apply to every side.

  • slamcharged word — charged word chosen to move you, not inform you
  • SLAMALL-CAPS — ALL-CAPS shouting raises volume, not substance
  • recklesscharged word — charged word chosen to move you, not inform you
  • devastatecharged word — charged word chosen to move you, not inform you
  • DEVASTATEALL-CAPS — ALL-CAPS shouting raises volume, not substance

Plainer rewrite

?
Built by the very same rules: swap loaded words for plain stand-ins, lower ALL-CAPS, and trim pile-on punctuation. It shows one calmer way to say the same thing — edit it freely.

+72 neutrality

Officials criticize plan that could harm families

State who did what, with what evidence — and let the reader feel their own way.

bias-mirror.log
$ bias-mirror analyze --local
tokens: 8   flagged: 3   findings: 5
checksum: 28989696282de793
neutrality: 28/100   verdict: loaded
stored: nothing — analysis runs on-device only
status: ok

Ask AI

AI can be wrong. Not medical, legal, or financial advice.

Bias Mirror is a thinking aid, not a verdict. It flags rhetorical framing, not truth or falsehood — read the source and decide for yourself.