Fact-checking is not reading an article and deciding whether it “seems right.” It is a repeatable process of breaking a piece into its individual claims and checking each one against a reliable source. This guide walks through that process step by step, so you can check an article methodically and leave a record a reader can trust.
Step 1: Isolate the claims
You cannot fact-check an article as a whole. You check its claims. Read the piece and pull out every statement of fact: figures and statistics, dates, names and titles, direct quotes, and cause-and-effect claims. Opinions and clearly framed predictions are not factual claims, but the facts they rest on usually are.
Claims to pull out
- Numbers and statistics, and what they measure
- Dates, names, titles, and affiliations
- Direct quotes and who said them
- Causal claims (“X causes Y”, “X led to Y”)
- Claims presented as established fact
Step 2: Find primary sources
For each claim, find the source closest to it. The original study, not an article summarizing it. The official record, not a blog repeating it. The dataset, not a chart built from it. A page that merely restates a claim is not a source; it is just another place the claim appears.
Reaching a page that returns a 200 status is not proof the claim is supported. Pages go stale, get rewritten, or never said what they are cited for. Read the source and confirm it actually contains and supports the specific claim.
When a claim traces back to a chain of citations, follow it to the end. Claims often drift as they are repeated, and the original frequently says something narrower than the version you are checking.
Step 3: Verify figures and quotes
Figures and quotes are where errors hide most often, because they are easy to copy and easy to distort. For a statistic, find who first published it, what it actually measured, and over what period, then confirm the article uses it in that original sense. For a quote, find where it was said, read the surrounding context, and confirm it was not trimmed into a different meaning.
| Claim type | What to check |
|---|---|
| Statistic | Origin, what it measured, time period, and whether it is used in context |
| Quote | Where it was said, the surrounding context, and whether it was trimmed |
| Date or name | The primary or official record, not a secondary retelling |
| Causal claim | Whether the source shows causation or only correlation |
Step 4: Handle what you cannot verify
Some claims will not check out. You may find a reliable source that contradicts the claim, or you may find no reliable source at all. Both matter. A contradicted claim needs a correction. An unsupported claim, one you cannot trace to any reliable source, should be treated as unsupported rather than assumed true. The honest move is to flag it and propose removing or qualifying it, with your reason.
Never fill a gap with an invented source or a number that “sounds about right.” If a claim cannot be supported, the correct outcome is to say so. Fabricating support is worse than the original error.
Step 5: Record what you checked
A fact-check is only as trustworthy as the record behind it. Note which claims you checked, what you found, and the sources you used. That record does two things: it lets the publisher act on specific issues, and it lets a reader trust that the check was real and thorough rather than a rubber stamp.
On CitePep, a fact-check becomes a public record: the claims checked, what was corrected, and the sources used, tied to the named person who did the work. That is what turns a fact-check into something a reader can verify.
Break the article into claims, trace each to a primary source, verify figures and quotes in context, flag what you cannot support, and record it all.
Common questions
What is the first step in fact-checking an article?
Isolate the checkable claims. Read the article and pull out every statement of fact: figures, dates, names, quotes, and causal claims. You cannot check an article as a whole, only its individual claims.
What counts as a reliable source?
A primary source closest to the claim: the original study, the official record, the person quoted, the dataset itself. A page that merely repeats a claim is not a source. Reaching a page with a 200 status is not proof; read it and confirm it actually supports the claim.
How do I check a statistic or figure?
Trace it to its origin. Find who first published the number, what it actually measured, and over what period. Numbers are often quoted out of context or drift as they get repeated, so confirm the original meaning, not just that the number exists somewhere.
What do I do when a claim cannot be verified?
Flag it. If you cannot find a reliable source, treat the claim as unsupported and say so, rather than assuming it is true. An honest correction proposes removing or qualifying the claim, with the reason.
Should I record what I checked?
Yes. A fact-check is more trustworthy when the record shows which claims were checked, what was found, and which sources were used. That record is what lets a reader trust the result rather than take it on faith.