Last updated: October 2026
Primary documents first, convergence never as proof, numbers interrogated — the Bureau's verification method, step by step.
A geopolitical claim can travel around the world before the evidence catches up. The Bureau’s method for checking one is the same method its research desk uses before publication: start from primary material, test the source, triangulate without fooling yourself, interrogate the number, and check the record. Five steps, in order.
Every article at the Bureau begins at the research desk with primary material: the official statement, the filing, the dataset, the treaty text, the market close. Secondhand accounts are a last resort, never a foundation. Apply the same rule to any claim you meet. Ask: what is the document behind this? If the claim quotes a leader, find the transcript or the official release. If it cites a dataset, find the dataset. If it cites a treaty, find the treaty text. If the source document cannot be found, the claim stays unverified — and the Bureau says so in the piece rather than printing it as fact. A claim with no findable source is not “probably true”; it is unproven.
The Bureau is a desk-based newsroom. It does not claim reporters in every capital, and it does not present secondhand material as field reporting. Hold every other source to the same standard. Was the outlet present, or is it relaying someone else’s account? Does the source have a direct interest in the claim — a government announcing its own success, a company defending its market, a faction shaping a narrative? Interest does not make a claim false, but it tells you how hard to check. Anonymous claims deserve the hardest check of all: if you cannot see who said it, you cannot weigh why they said it.
Before a major story reaches the page, the Bureau reads it through three lenses: how it looks from the West, from the East, and across the Global South. The desk stress-tests the draft from each vantage point and pushes back against each lens’s own assumptions. Do the same with any claim: find how it reads from at least two independent vantage points, not just two outlets repeating the same wire copy. And remember the Bureau’s warning: a fact is established through verifiable evidence, never through agreement. Three perspectives can share the same wrong assumption, and when they do, their convergence proves nothing. Where the lenses diverge, do not pick a winner and do not blur the difference: lay out who holds which view and what evidence each side is standing on.
A figure without an explanation is decoration. Every number in a Bureau story comes with its meaning in ordinary language: what it measures, over what period, compared to what. Apply that test to any statistic in a claim. “Exports doubled” — doubled from what, over what period, in value or in volume? “20% of oil” — 20% of what: global production, global consumption, seaborne trade? “The largest” — largest by what measure, and who measured it? If a statistic cannot be explained in one sentence, it does not belong in the piece — and it should not survive your check either. Charts deserve the same treatment: read the axes, the baseline, and the source line before you read the headline.
Finally, check what is already on the record. The Bureau publishes corrections on the page, with a date — it does not silently edit published claims. If a claim you are checking has already been corrected, the correction is part of the story. Use the Bureau’s own reference works: the Factbook’s country profiles for the standing facts, the Archive for what was reported and when, and the search box in the header to find whether the claim has been covered before and what the evidence then was. A claim that contradicts the standing record needs stronger evidence than one that fits it. And if, after all five steps, the claim still cannot be verified — do not run it. The Bureau’s rule is the reader’s rule too: when the desk cannot verify a claim, the claim does not run.
How the Bureau reports · Corrections Policy · Factbook · Archive
