Fact-checking is the step where every claim in a piece of content is confirmed against a reliable source before it is published. It covers statistics, dates, names, prices, legal statements, product details and quotes, and it ends with each claim either verified, corrected or removed.
How fact-checking works
A good check starts by listing the claims, not by rereading the article. Go through the draft and mark anything a sceptical reader could ask you to prove: a percentage, a deadline, a statement about what the law requires, a description of how a tool behaves. Opinions and advice do not need a source, but anything presented as fact does.
Each marked claim is then traced to the place it came from. The aim is the original source, not the blog post that repeated it. For UK figures that usually means:
- The ONS for population, employment, household spending and business statistics.
- GOV.UK for tax thresholds, deadlines, regulations and official guidance.
- Companies House for company names, registration details and filing history.
- The regulator itself (the ICO, FCA or ASA, for example) for what the rules actually say.
The checker records the source and date beside each claim, so anyone updating the page later can see where a figure came from. Anything that cannot be traced is rewritten as an opinion, softened to what can be proved, or cut.
Why it matters
An error costs more than an embarrassed correction. A wrong VAT threshold or filing deadline on an accountant’s blog can cost a reader money and the firm its credibility. A health or finance claim that cannot be backed up can bring a complaint to the Advertising Standards Authority, because UK advertising rules expect you to hold evidence for objective claims before you make them, and that includes claims on your own website.
There is a search angle too. Google’s guidance on E-E-A-T puts trust at the centre, and accuracy is the most basic form of trust. Pages that state things plainly and cite where the numbers came from are easier for a reader, a journalist or an AI system to rely on and quote.
Drafting with AI has made checking more important, not less. Language models can produce a confident statistic or a court case that does not exist, a failure known as hallucination. If AI touches a draft, every specific claim in it needs verifying by a person.
Common mistakes
- Citing a secondary article that quotes a study, without opening the study to see what it really found.
- Using US or global figures in a UK article as though they described the UK market.
- Leaving out the date. A statistic from several years ago may still be the latest available, but readers need to know its age.
- Checking the numbers but not the names, job titles, product features or prices, which go out of date faster.
- Treating fact-checking as a one-off. Rules and figures change, so pages that rely on them need a review date.
How to act on it
Build the check into your publishing process rather than relying on a careful writer. A simple routine works for most small businesses: the writer adds a source link or note beside every factual claim in the draft, a second person checks each one against the original, and anything unverified is removed before the page goes live.
Keep a short list of approved sources for your sector, so writers start from the ONS or GOV.UK rather than whatever ranks first in a search. For pages that depend on changing figures, such as tax rates, grant schemes or platform policies, add a review date to your content calendar.
When I plan and edit content as part of my content SEO work, sourcing is written into the brief, so the writer knows which claims need evidence before drafting starts.