Fact Checking Policy
How we verify our guides and keep the quality of community reports high.
Why This Matters
CallerCheck has two sides: editorial guides written by our team, and a phone number database built from community reports. Fact checking works differently for each. Our guides go through a proper editorial review. Community reports go through automated filtering and manual moderation. This page explains both.
Part 1: Editorial Content (Guides, Articles, Scam Alerts)
This section covers how we check the content our team writes and publishes.
What We Check in Our Guides
- Statistics and data: Any numbers or percentages we quote are verified against the original source.
- Regulatory information: References to UK law, Ofcom rules, or ICO guidance are checked to make sure they are correct and current.
- Scam alerts: Warnings about specific scam types are based on confirmed reports from official bodies, backed up by patterns in our community data.
- Contact details: Phone numbers, websites, and addresses for organisations like Action Fraud and the TPS are checked to make sure they still work.
- Practical advice: Any steps we tell people to take are checked against official guidance from the relevant body.
How We Check Our Guides
Go to the original source
We trace every factual claim back to where it came from. We always use the original source first, whether that is an Ofcom report, an Action Fraud bulletin, or a government publication. We do not rely on secondhand reporting when the original is available.
Verify the data
Statistics and figures are checked against the official dataset they came from. We note where the data is from and when it was published, so readers can look it up themselves.
Check the rules
Anything that references UK regulations, consumer rights, or official procedures is checked against the latest published guidance. Rules change, and we make sure what we are saying is still correct.
Second review
Someone else on the team reads the whole thing through, checking the claims, the sources, and the overall accuracy before we hit publish.
Sources We Use
Not all sources are equal. We rank them by how reliable they are:
Primary sources (we start here)
Official publications from Ofcom, Action Fraud, the ICO, HMRC, the National Crime Agency, UK Finance, and other government or regulatory bodies. This is where we go first and what we trust most.
Secondary sources (for context)
Reporting from established UK news outlets, telecoms industry publications, and recognised consumer bodies like Citizens Advice and Which?.
Community data (for spotting trends)
Patterns from CallerCheck user reports. We use this to spot new scam types early, but we always check against official sources before writing about it in a guide.
Part 2: Phone Number Data (Community Reports)
This section covers how we check the quality of community reports on phone numbers.
Automated Screening
Every community report is automatically screened before it goes live. The screening checks for:
- Racism and hate speech: Any racial slurs, discriminatory language, or references to ethnicity or nationality in a derogatory way. This includes common patterns like blaming scams on specific nationalities.
- Threats and harassment: Threatening language directed at the caller or anyone else.
- Personal information: Names, home addresses, account numbers, or anything that could identify a private individual.
- Spam and promotion: Advertising, promotional content, or links.
- Defamation: Unsubstantiated accusations that could damage a person or business beyond describing the call experience.
Reports that pass screening go live. Reports that get flagged are held back and queued for manual review.
What is allowed: Colloquial language and British slang are fine. Minor swearing in context is acceptable ("bloody annoying calls" is not going to get flagged). Spelling and grammar mistakes are not a problem. The test is whether the report is genuinely useful for someone trying to work out who called them.
Manual Review
Reports that get flagged by automated screening are reviewed by a member of our team. They can approve the report, reject it, or edit it to remove personal details before approving. Reports that are flagged by other users after publication also go into the manual review queue.
Community Voting as a Quality Check
Once a report is live, other users can mark it as helpful or unhelpful. This is not just cosmetic. Votes affect the risk score we calculate for the number. If the community consistently marks a scam report as helpful, it carries more weight. If a report gets voted down, its impact on the score is reduced.
This means the community itself acts as an ongoing quality check on the data. Misleading or unhelpful reports naturally get less influence over time.
How Risk Scores Work
We do not just count reports and call it a day. The risk score for each number is calculated using a weighted formula that considers the safety ratings across all reports, the total number of reports (more reports means higher confidence), and how the community has voted on each report.
A number with a single report is treated with less certainty than one with dozens. The score becomes more accurate as more people contribute. This means a single bad report cannot wreck a legitimate number's rating, and a single good report cannot disguise a known scam number.
Keeping Things Current
- Guides: Reviewed on a schedule. When rules change or new scam types appear, we update the relevant guides straight away and show the last updated date.
- Phone number data: Updated continuously as new reports come in. Risk scores recalculate automatically when reports are added or votes change.
- Fixing mistakes: When we find an error in editorial content, we correct it and follow our corrections policy.
Spotted Something Wrong?
If you notice an error in one of our guides, something out of date, or a community report that looks dodgy, tell us through our contact page. Send us the page URL and let us know what looks wrong so we can look into it.
We look at every report and aim to get back to you within two working days.