Trust FAQ

Haze indexes public publisher metadata, article URLs, source-review records, and approved provider data. New sources are matched to domains and reviewed before high-confidence labels are shown.

Haze can reference licensed or public rating sources such as NewsGuard, Ad Fontes, MBFC, and AllSides, while keeping provider attribution, version, effective date, and confidence visible.

Story clusters group articles by shared event, topic, entities, timing, and source context. Thin or uncertain clusters keep lower-confidence language so Haze does not overstate coverage patterns, and story pages disclose source count, article count, clustering confidence, and review caveats.

Haze stores source metadata, links, timestamps, and copyright-safe excerpts. Articles marked blocked or internal-processing-only do not expose restricted text in public reader surfaces.

AI summaries are limited to indexed coverage context, source mix, and visible caveats. Story detail labels the local model version, indexed-signal provenance, generated inputs, limitations, and whether external provider-backed attribution is attached.

Readers should send the Haze story URL, publisher URL, rating or attribution concern, and public evidence. Support routes correction requests to source review, story clustering review, copyright handling, or legal/privacy intake as appropriate.

Some panels still use indexed Haze signals rather than live provider feeds, including selected blindspot explanations, reading-history analytics, provider-backed AI attribution, and live newsletter/CMS issue content. Those surfaces label indexed provenance instead of implying external verification.