2026 Iran-Gulf Crisis Tracker
CC
Main
Live Data
Analytics
Tools
Reference

Trust

Editorial policy

What rules ConflictClarifier follows when deciding what to publish, how to label it, and when to correct it.

Editorial principles

  1. Distinguish report, claim, and verified fact. A government statement is a report. The thing it asserts is a claim. A claim becomes a verified fact only when independent evidence supports it. We label all three.
  2. Show the source list. Every event and claim links to the underlying sources. If we cannot, we say so explicitly.
  3. Be cautious about casualty figures. First-day numbers from any side — government, militant group, hospital, NGO — are routinely revised. We attribute the number to its source and update as figures converge.
  4. Don't launder propaganda.We include state-affiliated and partisan sources, but we label them and we don't describe their claims as facts.
  5. Show what we don't know.Confidence ratings, unverified flags, and the "data limitations" page exist to surface absence, not just presence.
  6. Correct in public. Errors are corrected with a visible note, not silently overwritten.

Neutrality

Neutrality on this site means neutral about framing, not neutral about evidence. If two governments make incompatible claims, we report both with their sources and verification status. If one claim is contradicted by satellite imagery, official rejection, or first-hand reporting, we mark it accordingly — neutrality does not require pretending evidence does not exist.

Handling official claims

  • Government and military statements are treated as primary sources for what that government is saying, not for what is independently true.
  • Official communiqués from multiple sides on the same incident are recorded separately, with their sourcing intact.
  • We do not promote unverified official claims to "verified" without independent corroboration, regardless of which government is making the claim.

Handling propaganda and information warfare

Propaganda is part of conflict. We document it rather than ignore it. State media, regime-aligned outlets, and overtly partisan publications appear in our source lists when they are first to report something — labelled clearly. We do not amplify atrocity claims, conspiracy theories, or doxxing material from any side. See the propaganda hub for definitions and examples.

Handling unverified claims

Claims that have not been independently corroborated receive an Unverified badge and a short note about what would change the rating. We publish them because their existence is part of the information record — but they are clearly marked, not treated as facts.

Conflict-sensitive language

  • We use the actor's self-identifying name where one exists and is unambiguous (e.g. "Hezbollah", "IRGC"), and add a neutral descriptor if the term is contested.
  • We avoid dehumanising language. Reports of mass casualties are written with the same restraint regardless of which side suffered them.
  • We avoid certainty language ("always", "never", "clearly the work of") where the evidence does not support it.

Civilian harm reporting

Civilian casualty figures are attributed to the body releasing them — health ministry, OCHA, ICRC, IDF, etc. — and not summed across incompatible sources. Where figures from different sources diverge by more than 20%, the divergence itself is noted on the event page rather than hidden behind a single number.

Casualty number policy

First-day casualty numbers are almost always wrong. Our policy:

  1. Report the number with its source and a low confidence rating.
  2. Hold off on dashboard-level aggregation for the first 24 hours where possible.
  3. Revise upward or downward as figures from independent humanitarian organisations converge.
  4. Keep the revision history visible on the event page.

Corrections

See the corrections policyfor how to report errors and how we respond. Corrections are logged with a visible note. Significant corrections (e.g. a verification status change) are surfaced in the "what changed" section of the next daily briefing.

Source attribution standards

  • Every event page lists its underlying sources with link and source name.
  • State media and partisan sources are labelled. See the sources page for the taxonomy we use.
  • Syndicated reports (same story from one wire on twenty outlets) count as one independent source, not twenty.

AI-assisted content

AI is used for narrow, well-defined tasks — generating bias framings of an event, summarising daily activity, classifying article topics. AI is not used to verify claims, declare facts, or replace human review on contested incidents. See the AI use policy for the full list.

Human review

High-impact items — casualty escalations, claims marked "verified", peace proposal status changes — are reviewed by a human before they are surfaced on the homepage. Lower-risk content (background articles, glossary terms, archival event ingestion) is editorially reviewed periodically rather than per-item.