Editorial Methodology

Research process

Each report begins with a geopolitical trigger from GDELT event monitoring or curated RSS feeds. A two-stage pipeline processes each trigger:

1. Geopolitical gate — a fast classifier rejects domestic crime, natural disasters without conflict angle, entertainment, and corporate news with no state actor involvement. Only cross-border incidents involving governments, militaries, or international organisations proceed.
2. Deep research — an AI research system executes targeted web searches (up to 5 queries per story), fetches full text from wire reports and official statements, identifies contradictions between sources, and synthesises a structured article. Searches target primary wire agencies first: Reuters, AP, AFP, BBC. Secondary sources (specialist defence and regional outlets) are used for context and corroboration.

Source hierarchy

Sources are weighted in three tiers:

• Tier 1 — Primary: Named wire agency reports with byline (Reuters, AP, AFP, dpa), official government or military statements with named spokesperson, official UN/NATO/ICC press releases. Two Tier-1 sources independently confirming the same specific fact = Confirmed.
• Tier 2 — Specialist: BBC, Al Jazeera, Financial Times, Defense One, The War Zone, Breaking Defense, Al Monitor, Jane's. One Tier-2 source = basis for credible reporting, not independent confirmation.
• Tier 3 — Aggregators and state media: MSN, Yahoo News, TASS, Press TV, Xinhua, SANA. State media from a party to the conflict = Unverified regardless of volume. Aggregators add no original reporting.

Wikipedia is not used as a primary source for contested military claims. Where Wikipedia is the only available background source, its cited primary reference is fetched and evaluated directly.

Claim classification

Every factual claim in a report is classified:

• Confirmed — Two or more genuinely independent Tier-1 or Tier-2 sources agree on the same specific fact. The same AP wire republished on 50 sites counts as one source.
• Contested — Sources explicitly contradict each other. Both claims are shown, with the exact source of each named. No editorial position is taken on which is correct.
• Unverified — Single source only, state media without independent corroboration, or anonymous sourcing ("officials said"). The reason for unverified status is stated explicitly.

Confidence scoring

Every report carries a confidence score from 0 to 95. Scores are never 100 — certainty is not achievable in conflict reporting.

• Base score: 50
• +10 per genuinely independent source (maximum +30)
• +10 if a named primary source with byline is present
• −10 per contested claim in the report
• −20 if all sources are state media

A score of 70+ indicates high confidence: multiple independent sources confirm the core claim with at least one primary source present. A score below 50 indicates significant uncertainty: the event is reported but the details are contested or from a single source.

Synthesis standard

Every report is required to contain information that goes beyond what any single wire service published. This information gain may take the form of:

• Explicit identification of contradictions between named sources
• A confidence-weighted synthesis across multiple conflicting accounts
• Second-order effects: economic, humanitarian, or diplomatic consequences identified from specialist sources
• Historical context connecting the current event to prior incidents in the same conflict

Reports that only restate what wire services already published are not published. The synthesis layer is the minimum standard for publication.

Corrections policy

Errors in published reports are corrected as soon as they are identified. A correction updates the article with a visible note stating what was changed and why. The original erroneous text is not silently deleted.

To report an error: identify the specific claim, the report URL, and the source that contradicts it. Corrections based on primary wire sources or official statements are prioritised. Corrections based on social media or anonymous claims are not accepted without independent corroboration.

AI and automation disclosure

Reports on Conflict Updates are researched and drafted using AI (Claude by Anthropic) with live web search capabilities. The AI system searches primary wire sources, fetches full article text, identifies source contradictions, and produces a structured article following the methodology described on this page.

All published content is reviewed against stated source attribution. The confidence scoring, claim classification (Confirmed / Contested / Unverified), and source hierarchy are applied programmatically per the rules above — not editorially adjusted after the fact.

Conflict Updates does not publish AI-generated speculation, prediction, or opinion. Every claim in a published report has a named source.