Methodology

Source tiers

Every stream that backs an event is tagged with a tier:

• T1 — Major outlet. Pre-registered news organisations with editorial standards (BBC, Reuters, CNN, Al Jazeera, AP, others).
• T2 — Verified. YouTube-verified channels with significant audience and a track record we've confirmed.
• T3 — Default. Everything else, weighted by the channel's observed reliability score.

Event clustering

Streams about the same real-world event are grouped using a combination of fuzzy title matching, named-entity overlap (people, places, organisations), and temporal proximity. The same tornado outbreak covered by ten chasers becomes one event with ten sources, not ten events.

Claim consensus

When sources make specific claims (“casualty count”, “suspect identified”, “evacuation ordered”), we score them by how many independent sources confirm the same fact:

• Confirmed. 4+ confirming sources.
• Developing. 2-3 confirming sources.
• Unconfirmed. Single source.

Contradictions are flagged when sources make mutually-exclusive claims about the same fact. The API returns both the consensus statement and the contradiction map.

Reliability score

Each channel has an observed reliability score that updates over time as their claims are confirmed or contradicted by other sources. Channels that consistently surface accurate, early reporting accumulate score; channels whose claims are repeatedly contradicted lose score. The score weights how heavily a T3 channel's claim contributes to consensus.

What we don't do

• — Editorialise. The API surfaces consensus and contradiction; it does not adjudicate truth.
• — Hide the disagreement. If sources contradict, the contradiction map is visible.
• — Pretend to be exhaustive. We index live streams from major platforms; we don't claim every event everywhere.