Fact-Checking Policy
This page describes how a claim gets from somebody’s mouth, or out of a PDF, into a sentence we publish, and what has to happen to it on the way. Last updated: August 2026.
Who checks what, before anything publishes
The writer checks first and files with the sources attached, so every figure, quote and assertion traces to a link, a document, a recording or a note of who said it and when. A draft arriving without that goes back before anybody reads it for style. The section editor then reads for the claims and a second editor reads for the assumptions, which rarely catch the same problem.
A second set of eyes on the verification itself is mandatory rather than discretionary in these cases.
- A person or a company is accused of something.
- A number is the story rather than the furniture around it.
- The piece quotes or characterises a document.
- A study or a dataset is the basis of the claim.
- Everything rests on a single source.
- The story is still moving and we are publishing inside the first hour.
The explanation gets checked as hard as the news
The explanation is the product here, so it is verified like any other claim. A definition comes from the body that owns it: the competition’s own regulations for squad cost rules, the government’s own notice for a visa category, the manufacturer’s filing for a recall turning on battery chemistry. Another outlet’s summary of that rule is a lead worth following, not a source we can stand behind.
Then two further tests, both of which send drafts back. Is the definition current, because rule-based explanations go stale quietly and a paragraph that was right in March can be wrong by August. And does it hold without leaning on a second term the reader also does not know.
Primary sources, and what a statement is evidence of
We would rather read the ruling than a report of the ruling. Filings, official notices, transcripts, published datasets and on-record interviews come first, and secondhand accounts are used where there is nothing better and labelled as what they are. If we could not get hold of the document, the copy says we were told about it.
A company statement or a ministry press release is a primary source for one thing, which is what that organisation chose to say. It is not evidence that the thing happened. We quote it, attribute it, and then go looking for whether the claim underneath holds, which on a busy day is the step most likely to get skipped.
Numbers, studies and surveys
A figure is checked against the original data and not against the summary that carried it to us. We look at what was measured, over what period, and whether the finding supports the sentence somebody wants to write on top of it. Sample size matters. Who paid for the work matters. Whether a paper went through review or sits on a preprint server is said in the copy.
Surveys need a sample, field dates and a named commissioner before we will use them, and where we cannot get the methodology we do not publish the number at all. A percentage with no method behind it is a shape, not a fact. When two official bodies publish different numbers for the same thing, we report both and name both rather than splitting the difference.
Pictures, video and anything we did not shoot
Every image from outside gets a reverse image search, a look at whatever metadata survived the upload, and an attempt to reach whoever posted it first. Where the location is load-bearing we match visible features against maps and street imagery, and check that the light and weather fit the time being claimed. Nothing runs if we cannot establish where it came from.
Generated and manipulated imagery is part of this check now rather than an occasional curiosity. We do not illustrate a real event with a synthetic image, our own illustrations are labelled as illustrations, and studio publicity material is labelled a handout, because a free photograph is doing a job for whoever handed it over. What our tools are used for is written out in our editorial standards.
Social posts and material sent in by readers
A checkmark on an account verifies a payment, not a fact. Before we cite a post we try to reach the person who made it, establish they are who they say and were where they claim, and ask whether the file is theirs or something they found.
We do not report a viral claim as news on the grounds that it went viral, which is a way of publishing something false while technically only describing it. Where a post has since been deleted we archive it, describe what it said, and tell the reader the account removed it.
Publicists, planted stories and access
Entertainment coverage arrives pre-spun and we treat it that way. An unnamed source close to a performer is refused as a category unless we know exactly who it is and the reason for withholding the name meets the test on our standards page, because that phrase nearly always describes somebody paid to prefer one version of events.
Access is not traded for coverage. We honour an embargo we agreed to, we decline embargoes arriving with conditions attached to the framing, and we accept no quote approval. Junket access withdrawn over something we published becomes part of the story. Statistics inside a press pack get the same methodology treatment as any other number.
The same standard, whoever is speaking
A claim is checked against the evidence rather than against the person making it, and a politician who has been reliable before earns no lighter check. We do not set a verified claim beside an unverified one for the look of even-handedness, because that is a way of moving a false statement into print behind somebody else’s name. Where the evidence does not settle a question, we say we could not establish it.
We work to the principles the International Fact-Checking Network sets out on non-partisanship, transparency of method, and open corrections. We are not a signatory to anything and do not claim to be. When verification fails after publication, which it does, our corrections policy governs what happens and who decides. A reader who thinks a check went wrong can write to support@significadopedia.com and reach an editor.
