Rule 01
Start with a claim, not a side
Write the proposition in testable language before collecting evidence. The conclusion is allowed to disappoint any audience.
Editorial method
Neutral rules. Evidence-weighted conclusions.
Being nonpartisan does not mean placing the truth halfway between two claims. It means testing every claimant by the same visible method.
Rule 02
Use the closest source
Prefer datasets, legislation, archives, regulators and original research over summaries. A reel or podcast can reveal a question; it cannot prove the answer.
Rule 03
Define every important number
Record the measure, geography, timeframe, denominator and revision status. Visa grants, arrivals and net overseas migration are different things.
Rule 04
Test causation
Sequence, mechanism, scale, alternatives and counterfactuals matter. Correlation alone is not a causal result.
Rule 05
Make uncertainty visible
Say what the evidence cannot establish and what new evidence would change the conclusion.
Rule 06
Correct the record
Material corrections are dated, explained and attached to the article rather than quietly overwritten.
Rule 07
Maintain the answer
Every article records whether its evidence is current, historical or developing, plus the date or event that triggers its next review.
Evidence threshold
The 90-point publication gate
High-stakes work must pass source quality, claim coverage, causal reasoning, uncertainty, fairness, harm minimisation, clarity, link durability and correction-readiness checks. A strong headline cannot compensate for a missing evidence chain.
AI use
Tools do not become sources
Automation may help organise material, but every substantive claim must resolve to evidence a reader can inspect. Generated quotations, citations or identities are prohibited.
Learning system
Data can improve the work without deciding the truth
Research findings, corrections and qualified reader behaviour may improve questions, explanations, formats and timing. They cannot weaken evidence, fairness, privacy, harm controls or the correction standard. Material changes are recorded with a success measure and review date.