Verification Methodology — Sentinel Review

Confidence levels

● Verified

The event is corroborated by at least two independent sources from different platforms or organisations, and at least one of the following: geolocated footage, official acknowledgment, or matching local press wire.

≥2 independent sources, different platforms
+ geolocation OR official acknowledgment OR wire corroboration

◐ Partial

Multiple sources exist but all are from the same platform, or a single verified-tier source is paired with one corroborating circumstantial signal. The event likely occurred but cannot be fully confirmed.

≥2 sources, same platform
OR tier-1 source + one corroborating signal

○ Unconfirmed

A single source, or multiple sources that trace back to a common origin (e.g. all citing the same Telegram channel). Published for awareness; treat all unconfirmed events with significant caution.

Single source
OR multiple sources with common origin

High-impact hold: Events flagged by the extraction pipeline as involving mass casualties, nuclear or chemical signals, or significant escalatory steps are held in the human review queue regardless of source count. They do not appear on the live map until a human reviewer approves them.

Source trust tiers

Every source in the system is assigned a trust tier by a human editor. Tiers affect how much weight a source's posts receive in confidence scoring — a tier-1 source posting alone can reach partial confidence; a tier-3 source posting alone stays unconfirmed.

Tier 1

High trust

Verified investigative accounts with established geolocation track records, major international wire services (Reuters, AFP, AP), and peer-reviewed OSINT projects with transparent methodology. Examples: @DefMon3, @OSINTtechnical, Reuters Wire.

Tier 2

Medium trust

Established mapping or conflict accounts with a good verification track record but less transparent methodology, regional press with known editorial stances, and community geolocation projects. Require corroboration before reaching verified.

Tier 3

Low trust

Anonymous accounts, state-affiliated media, single-contributor Telegram milblogs, and sources with known track records of amplifying unverified or false information. High volume but low signal. Never sufficient alone for any confidence level above unconfirmed.

Event processing pipeline

Ingestion

Posts are collected from curated source feeds (X, Telegram public channels, RSS, wire services) at regular intervals. Every post is stored verbatim in an append-only log and never mutated.

Entity extraction

An LLM (Claude) analyzes each post and extracts: event type, location, approximate time, actor (if identifiable), and a brief description. Posts without a discrete conflict event are discarded. Extraction output is structured JSON — the model cannot free-text outside the schema.

Deduplication and clustering

Candidate events within 15 km and 2 hours of each other are clustered as potential duplicates. A second LLM pass decides whether two candidates describe the same event or distinct events. Merged candidates inherit all source links.

Confidence scoring

Each event is scored using the rules above. Source trust tiers are applied as weights. The confidence label shown on the map is deterministic from the rule set — not a probabilistic output of a model.

Human review queue

High-impact events and events with a contradicting source are held for human review before publication. A reviewer can upgrade or downgrade confidence, add notes, or reject the event entirely. The reviewer's decision and notes are published alongside the event.

Content and graphic material

Sentinel Review does not embed media directly. Source links open in a new tab and may contain graphic content including footage of combat, casualties, and infrastructure destruction. Every source link is accompanied by a content warning.

Screenshots of source posts are stored in object storage for archival purposes (in case the original is deleted or modified) but are not displayed by default. They are available on the event detail page behind an explicit expand action.

Errors and corrections

Every event page has a “Report this event” link. If you believe an event is incorrectly located, mislabeled, or based on fabricated sources, use that link. Reports go directly to the editorial queue and are reviewed within 24 hours.

Corrections are published openly in the event's change history. Confidence can only go down, not up, once a credible dispute has been filed.

To report an event error: use the “Report” link on the event page, or email corrections@thesentinelreview.com