The fact-check toolkit: how to test any claim yourself

You will meet many more claims than this site can check. This page gives you the method, the sources and a worked example so you can do it without us.

Last reviewed: 2026-11-06

Five questions for any claim

1. Who says it, and what do they gain? An agency that says “our women are all verified” sells verification. A forum poster who says “I know a hundred happy couples” may be selling a membership. This is not a reason to dismiss a claim, only a reason to check it.
2. What is the primary source? A statute, a government page, a dataset, a peer-reviewed study. A blog that quotes another blog is not a source. Follow the link until you reach a document that was produced by the people who measured or wrote the rule.
3. What are the date and the denominator? “Divorce rates are high” needs a year, a country and a base: divorces per what? A number without a denominator is decoration.
4. Does the comparison hold? Are the two groups defined the same way, in the same year, by the same method? A claim that compares a Russian average with an American “family” figure is usually comparing two different things.
5. What would change my mind? If nothing could, you are not checking, you are shopping for support. Write down what evidence you would accept before you look.

Primary sources worth bookmarking

Worked example: “Russian women outnumber men”

We will run the five questions on a claim you will hear constantly.

1. Who says it? Often an agency, to suggest that women have few options at home.
2. Primary source? The share of women in the population is a standard indicator: the World Bank publishes it as series SP.POP.TOTL.FE.ZS. You can retrieve it yourself from the API at https://api.worldbank.org/v2/country/RUS/indicator/SP.POP.TOTL.FE.ZS?format=json&mrnev=1, where mrnev=1 asks for the most recent non-empty value.
3. Date and denominator? The latest value we retrieved on 6 November 2026 was for 2025: about 53.6 % of the Russian population, as a share of the total.
4. Does the comparison hold? The same series gives about 49.8 % for the United States and 50.7 % for the United Kingdom, so the comparison is like for like. The gap is a few percentage points, not a landslide.
5. What would change my mind? The claim as stated (“more women than men”) is simply true. The claim people attach to it (“so they are eager to leave”) needs a different kind of evidence, such as surveys on motives, and the World Bank does not provide that.

Two exercises with real numbers

Practice on figures you can open yourself. Both come from the World Bank series used elsewhere on this site, retrieved on 6 November 2026.

Exercise 1: the life-expectancy gap

In 2024, life expectancy at birth was about 78.9 years for Russian women and 68.3 for Russian men. Subtract: the gap is about 10.6 years. The same subtraction for the United States (81.4 and 76.5) gives about 4.9. What can you conclude? That men in Russia die, on average, much earlier than women, and much earlier than American men. What can you not conclude? Anything about marriage, motives or character. The first statement is a measurement, the second would be a story. Learn to notice the moment a sentence moves from one to the other.

Exercise 2: two fertility rates

Russia’s total fertility rate was about 1.34 in 2004 and about 1.42 in 2024; the United States stood at about 2.05 and about 1.63. Notice that the Russian figure rose slightly while the American one fell by more than a third of a child per woman. A claim written in 2004 that “Americans have many more children” was roughly right then and is much less true now. Dates matter, and a source that does not give one is not a source.

What counts as a primary source

Rank what you read. At the top: the statute or dataset itself. Next: an official explanation of it, such as a government agency page. Next: a reputable reference that cites the first two. At the bottom: anything that does not name its sources, including most forum posts and agency blogs. Use the lower tiers to find leads, and the upper tiers to settle questions.

When sources disagree

Two reputable sources will sometimes give different numbers for the same thing. Before you decide one is wrong, look for three common causes: a different year, a different definition (births per woman versus births per thousand people, for instance), or a different method of estimation. Statistical agencies revise figures as new censuses and surveys arrive, which is why we always give the retrieval date and the database’s own update date. When the difference remains, report the range rather than choosing the number that suits you.

The same attitude applies to law. Statutes are amended, agency pages are reviewed on their own schedule, and a page marked as reviewed years ago may be stale. Treat the review date as part of the fact, and check the live page before you act on it.

Testing what an agency says

Agency pages are full of figures. Here is how to turn each phrase into a question you can put in an e-mail.

A legitimate provider can answer these questions in writing without irritation. An evasive answer is an answer.

Four traps in statistics

• The average hides the spread. “The average Russian family has one child” is a statement about an average, and no family is average. A fertility rate of about 1.4 is compatible with plenty of families with three children.
• Survivorship. Testimonials come from couples who stayed together and wanted to be quoted. The ones who did not stay do not write reviews.
• Old data, presented as current. A claim from 2004 may have been wrong then and is certainly dated now. The archive articles we restored on this site carry their original text and a correction note where it is needed. See, for instance, the 2007 essay on agency advertising.
• Cherry-picked comparison. A choice of year or country can make almost any claim look true. Ask for the series, not the sentence.

One-page checklist

1. Write the claim down in one sentence, with a number if it has one.
2. Find the primary source and note its date.
3. Check the denominator and the comparison group.
4. Look for the same figure in a second, independent source.
5. Decide what you will do differently depending on the answer. If nothing, the claim did not matter.

Sources and date consulted

1. World Bank, World Development Indicators API, series SP.POP.TOTL.FE.ZS. Retrieved 2026-11-06 (database last updated 2026-10-08).
2. 8 U.S.C. § 1375a and § 1184(d), Cornell LII, consulted 2026-11-06.
3. USCIS, K-1 page, reviewed 2018-03-23, consulted 2026-11-06.
4. US Federal Trade Commission, ReportFraud.ftc.gov, and FBI Internet Crime Complaint Center, both consulted 2026-11-06.

Have a claim you would like checked? Send it to us with a link to where you saw it.