← 2026-08-312026-09-01

Manipulation techniques in today's news

14 articles, drawn evenly from outlets rated left, center, and right. Each one lists the manipulation techniques our model found — not a judgment of whether it's true.

⚠️ This day is not evenly balanced: left 5, center 4, right 5. We had fewer articles available from center-rated outlets than from the others — usually because their feeds had already rotated the day's stories out by the time we built this page.

Can we afford Iran?

Loaded Language Appeal to Fear Exaggeration/Hyperbole
Vox · left 12.1% of sentences flagged

Get right out of town

Name Calling Appeal to Fear Loaded Language
Fox News · right 4.7% of sentences flagged

← Previous day (2026-08-31)

How these articles were chosen

  • Sources are a fixed, version-controlled list of outlets, not picked per day.
  • Equal numbers are taken from left-, center-, and right-rated outlets, using the AllSides media bias ratings — those labels are theirs, not ours.
  • Within an outlet we take the most recent articles from its politics or world section feed. No human picks the interesting ones.
  • Live blogs, podcasts, video and photo galleries are skipped — they contain too little text to analyze.

The full source list

Ratings below are AllSides' own labels (retrieved 2026-07-31), not ours. We group Left and Lean Left together, and Lean Right and Right together, which balances direction but not intensity — the exact ratings are here so you can judge that yourself.

  • Left: The Guardian (US) (Left), NPR (Lean Left), Vox (Left), Mother Jones (Left)
  • Center: BBC News (Center), Christian Science Monitor (Center), The Hill (Center), UPI (Center)
  • Right: Fox News (Right), New York Post (Right), Washington Examiner (Lean Right), Reason (Lean Right)

Limitations worth knowing

  • Semblen detects manipulation techniques, not left/right tilt.
  • The model has known false positives on strong-but-legitimate language — particularly military reporting and opinion writing in the author's own voice.
  • We list the techniques found. A technique flag is not a claim that an article is false, and we don't rank outlets against each other.
  • Articles where the model found nothing are not published — there'd be nothing on the page.
  • Each article is analyzed once, at the time shown on its page. Outlets sometimes update articles after publication, so a quoted phrase may no longer appear in the current version.
  • Findings may include language an outlet is quoting rather than asserting — a politician's attack line, or an opponent's own words. We flag technique wherever it appears, including inside quotations.
  • We link to each original article and quote only short phrases as evidence.

Get notified as this improves — new detection models, the browser extension, and outlet scorecards. Occasional email, no spam.