Fact-Checking Policy
This page describes how a claim gets checked before it appears, in the order the checks happen. Last updated: August 2026.
Our format concentrates the risk in one place. A conventional news story spreads its facts over twenty paragraphs. We put the answer in the first two sentences, which is where most readers stop, so most of what follows is about that sentence.
The checking starts before the writing does
An assignment leaves the desk with the question in the reader’s own phrasing and a note on what would count as an answer: which document, which body, which dataset. Settling that in advance stops a writer gathering whatever happens to be available and calling the result an answer.
The writer verifies each claim as they go and keeps the source attached, so the draft reaching the desk has a link, a document reference or an interview date against every number. A draft with an unsourced figure goes back unread. That is not pedantry. It is the only way an editor can check anything in the time there is.
What the second editor does
Every story is read by an editor who did not write it. The first pass is the house check: find the question, find the answer, confirm the answer is in the first two sentences. The second pass is the facts, worked from the writer’s sources rather than the writer’s confidence, and that editor opens the links.
Some stories need a third read, and the list is fixed. Anything alleging wrongdoing by a named person or company. Anything a reader could act on medically or financially. Anything about a recall or a safety notice. Anything resting on an unnamed source. Anything we are publishing faster than we are comfortable with. Those go to the managing editor first, and on a developing story that read is the step most likely to make us late. It is not skippable.
Primary sources, and what counts as one
The document, the dataset, the filing, the notice, the transcript, the person who was there. Not a rival’s write-up of any of those, and not a press release summarising them. Where the primary source is public, we read it and say we read it.
Secondhand reporting can start a story and cannot finish one. If we cannot reach the primary source in time, the story says which claim rests on another outlet’s reporting and links it. When the primary source is a company describing its own conduct, it is a primary source for what the company says, not for whether the thing is true.
Links, so you do not have to take our word for it
Claims are linked in the sentence that makes them, to the source itself rather than to a homepage or a search result. Where the source is a document, we link the document and not a page describing it. If it sits behind a paywall or a login, we say so before you click.
Where a source cannot be linked, because it is a paper record, an interview or a document given to us on terms, the story says what kind of source it is and when we saw it. Some of our claims cannot be checked from your chair. Better to label those plainly than to dress them up with a link to something adjacent.
Numbers, studies and the distance between a finding and a claim
Statistics are traced to whoever produced them and read in the original release rather than in somebody’s coverage. We check what was measured, over what period and against what baseline, because most misleading figures are technically accurate and quoted in the wrong units. Where official bodies publish competing numbers for the same thing, we give both and attribute each.
On research we read the study. Sample size, whether there was a control group, whether it was done in humans, whether it is a preprint, who paid for it, and whether the finding supports the claim being hung on it. A lot of health and consumer copy is an accurate report of a press release about a study of forty people. Where the evidence cannot carry the answer a reader wants, the answer is that it cannot.
Images, video and things people posted
Pictures are checked before they run. Reverse image search first, then whatever the file itself gives up: creation date, metadata, and what the visible detail says about the place. A photograph from a different incident three years ago has been offered to us more than once, usually by somebody entirely sincere.
Material from social accounts is a lead, not a source. We try to reach whoever posted it and check it against something we can verify independently before citing it. If we run it without reaching them, we say so. A pile of anonymous complaints about a product is a reason to start asking a manufacturer questions, not a fault we can report as established.
The same standard whoever is making the claim
A claim is checked the same way regardless of who said it, which party they belong to and whether they buy advertising from us. We do not wave a claim through because it came from an official podium, and we do not hold one to a higher bar because we find the person disagreeable.
Balance is not the goal, and false balance is one of the ways to fail at this. Where the evidence points one way we report that it does, and say what the other side offered in place of evidence. We work to the IFCN principles on non-partisanship, transparency of sources and open methodology. The rules the whole desk works to are in our editorial standards.
When a check fails after publication
Sometimes we clear something that turns out to be wrong, and the part that matters is what happens next. The story is corrected on the page with a dated note under our corrections policy, and the note says what was wrong rather than that the piece has been amended.
Then we look at the check that let it through, because a wrong fact usually means a step was skipped. If you think something we published does not stand up, write to support@faq-ans.com and tell us which sentence and why. Readers catch a fair share of what we get wrong, which is one reason the day starts with the inbox, as our about page explains.
