Fact-Checking Policy

Why this page exists

A general news site lives or dies on whether its facts hold. This page describes how a claim is checked before it reaches you, in enough detail that you can judge the method rather than take our word that we have one. It is the companion to our editorial standards, which set the principles; this page is about the practice of verifying a specific claim on a specific day.

Fact checking here is not a separate department that stamps a finished article. It is something every reporter and editor does continuously, before publication, as a normal part of writing the story. The checks below are the ones that get applied whether the piece is about an election, a company’s results, a medical study or a sports transfer.

The rule underneath everything: go to the primary source

The core of our method is simple to state and slow to do: we verify a claim against the primary source, not against another outlet’s account of it. If a report says a court ruled a certain way, we read the ruling. If a headline says a study found something, we read the study, not the press release about the study. If an official is said to have announced a figure, we find the announcement, the filing or the recording and check the number ourselves.

Secondary sources, meaning other people’s reporting, are useful for finding a story and useless as proof of it. A claim that has been copied across twenty sites is not confirmed twenty times. It is usually one unchecked source copied twenty times, and the more places it appears, the more it looks true without being any truer. We chase claims back to where they started, and if the trail ends at nobody who actually knows, the claim does not run as fact.

How a claim gets checked

The steps vary by story, but the questions do not.

• Who is the origin? We identify the first party who actually knows the thing, and distinguish them from everyone repeating it. A named, accountable origin is worth more than a viral one.
• What is the evidence? A document beats a description of a document. A dataset beats a summary of it. We look at the thing itself and quote it accurately, in context, not cropped to change its meaning.
• Does an independent source agree? For a significant or contested claim, we look for a second source that knows independently, rather than one that read the same first report.
• What do the numbers say? We check the arithmetic, the units, the date range and the baseline. A percentage with no base, or a record with no comparison period, is a red flag, not a headline.
• Who benefits from us believing it? A claim that conveniently serves the person making it gets more scrutiny, not less.

Numbers, studies and statistics

Data stories are where confident reporting goes wrong most often, so they get particular care. A single study is a data point, not a verdict, and we say how large it was, who ran it, whether it has been reviewed by other experts, and what its authors themselves say it does not show. We are wary of a dramatic finding that no one else has reproduced, and we distinguish a correlation from a cause, because a great many headlines do not.

With official figures we note the source and the period, and we resist the urge to turn a provisional estimate into a hard fact. Where we cannot fairly summarise a number without stripping the caveat that makes it meaningful, we keep the caveat, even at the cost of a cleaner sentence.

Images, video and things that spread online

Material that arrives through social feeds is guilty until verified. Before we report on a photo or a clip we try to establish where and when it was actually taken, whether it shows what it claims to show, and whether it is old footage recycled for a new event, which is one of the most common forms of misinformation. An image that cannot be stood up is not used as evidence of anything, and if we report that something is circulating online, we say plainly whether it has been verified or not.

When we are not sure

Sometimes the honest state of a story is uncertainty, and we say so rather than manufacture a confidence we do not have. We will tell you what is known, what is claimed but unconfirmed, and what is still unknown, and we will attribute contested claims to whoever is making them instead of adopting them as our own. On a developing story that uncertainty shifts by the hour, and we timestamp and update accordingly.

Refusing to state something as fact is not fence sitting. Reporting that one side has evidence and the other does not is a finding, and we report it. What we will not do is split the difference between a true claim and a false one and call the average balance.

When a check fails after publication

No process catches everything, and a claim that passed at the time can prove wrong later. When that happens we treat it as a correction, openly and on the original story, under our corrections policy. If you can see a claim of ours that does not hold against its primary source, you are doing our job with us, and we want the message: write to support@broadbrowse.com with the link, the claim and the source that contradicts it.

Last updated: August 2026