Methodology

Counting coverage

Every percentage on this site is computed from the source list shown on the story page. Each outlet that covered the story is placed in exactly one bucket — World (Left), Christian sources, or World (Right) — and the bar is those counts divided by the total, rounded so the three segments always add to 100%. Only classifications an editor has approved are counted. Anything pending, rejected or mixed is never drawn as a segment; it is shown next to the bar as "+N unclassified". If nothing has been approved yet, the bar is empty and says so rather than showing a guess.

Classifying an outlet

Classification is per story, not permanent. The bucket reflects how that outlet framed this particular story. A religious publisher counts as Christian even if it also leans politically, so nothing is double-counted. Each classification stores the evidence and the reason, both visible on the story page.

Writing each perspective

The left, biblical, and right versions are generated by AI from the story's evidence bundle — the headlines, the short snippets the outlets themselves publish in their feeds, and the metadata across every outlet that covered it. ALTARED never fetches, stores or republishes full article bodies, so the evidence is deliberately limited. When the evidence is too thin for the requested length, the text says coverage is insufficient instead of padding. The model is instructed never to add names, numbers, dates, quotes or outlets, and to keep contested claims marked contested. Versions are cached against the article's evidence revision and the active editorial policy — if either changes, the cached version is discarded and rewritten.

Length and perspective are separate

Choosing a 5-minute read does not change your perspective, and switching perspective keeps your chosen length. Each combination is stored on its own.

Audio

The listen button speaks exactly the version on screen using AI text-to-speech. It is a synthetic voice, labeled as such, and it never reads text that isn't shown.

Scripture and prophecy

Passages are quoted from the World English Bible, a public-domain translation, together with their surrounding context so a verse is never left floating. Connections are labeled by type: direct reference, ethical application, thematic parallel, or prophecy interpretation. ALTARED never asserts that a prophecy has been fulfilled; where readings differ, the competing views are printed alongside.

Where the news comes from

ALTARED monitors a registry of official published feeds from world outlets (BBC News, NPR, The Guardian, Fox News, Al Jazeera) and Christian outlets (The Christian Post, CBN News, Christianity Today, International Christian Concern, Open Doors), plus a Google News topic feed used only to check coverage breadth. The registry is editable in the newsroom, every feed's health is recorded on each run, and a feed that is dead or blocked is flagged and skipped — never simulated and never filled in with invented items.

Items are de-duplicated by link and by matching headlines, and syndicated copies collapse to the originating outlet. A cluster is published automatically once at least two outlets have covered it; single-outlet items wait for an editor.

Underreported

A story is marked underreported when fewer outlets than the configured threshold covered it in a trailing seven-day window across our monitored registry. That is a measurement of our own list, not a claim that anyone suppressed anything.

The briefing email

Sign-up uses double opt-in. No sending service is switched on for this project yet, so an address you submit is stored and no email is sent. We will never say a message went out when it did not.