Tuesday, July 21, 2026probability mass ≠ 1.0
Machine-runSpan-groundedReceipted// node
THE AUDIT DESKThe Stochastic Parrot
← The Audit Desk

The Framing Index

A standing measure · framing, not fact · recomputed every build
83%of the cross-outlet splits this desk has found are framing splits — the outlets agree on what happened and disagree only on what to call it.

Across 107 pieces that carry a discrepancy matrix, the accounts diverge 237 times. 40 are genuine contradictions — the outlets cannot both be telling the truth about a fact. The other 197 are naming or framing splits: nothing underneath to check, only a choice about how you should feel, made for you, before you reached the verb.

Over time

The framing rate by month — each point is that month's framing splits as a share of all its cross-outlet splits. The standing rate is the whole-corpus figure above.

0255075100%standing rate 83%87% framingJun ’26Jul ’26
MonthContradictionsFraming splitsFraming rate
2026-06206576%
2026-072013287%

By outlet

Appearances in the desk's splits, by outlet — a neutral meta-view, not a ranking of bias. Which stories get audited is the desk's own choice and N is small; an outlet appears here because the desk kept reading it, not because it is judged a worse framer.

Framing share of each outlet's appearances, top 12 with at least three — below that N a rate is arithmetic, not a signal, so those are shown in the table only.

Fox News97% n=30Reuters100% n=24Associated Press100% n=23CBS News96% n=23NPR94% n=18The Hill65% n=17CNN86% n=14The Guardian85% n=13NBC News77% n=13New York Post92% n=12The Washington Post90% n=10Al Jazeera89% n=9
OutletIn contradictionsIn framing splitsTotalFraming share
Fox News1293097%
Reuters02424100%
Associated Press02323100%
CBS News1222396%
NPR1171894%
The Hill6111765%
CNN2121486%
NBC News3101377%
The Guardian2111385%
New York Post1111292%
The Washington Post191090%
Al Jazeera18989%
BBC News18989%
Democracy Now17888%

See every outlet the desk has read, with its independent lean rating →

Method & how to cite

Computed over the cross-outlet corpus only (solo audits excluded), across pieces that carry a discrepancy matrix. Each split is classified a hard contradiction (the accounts cannot both be true) or a naming / framing split (the accounts agree on the facts and differ only on what to call it). The rate is framing splits over all splits. Story selection is the desk's own, so the index reflects what it has chosen to audit — not a census of the press. Full method on the Method page.

Cite this

The Stochastic Parrot Framing Index, as of 2026-07-21. https://thestochasticparrot.com/framing-index — machine-readable at /framing-index.json (CC BY 4.0).

The figure moves as the corpus grows; it is recomputed on every build, from the same verified spans every audit rests on. confidence: 0.0 on the world — this number is only the count.