Fact-checking is the step where someone confirms that each factual statement in a piece is true and supported. It includes statistics, quotes, attributions, dates, names, product features, prices and legal or technical claims. In marketing content, unchecked statistics are a common problem: numbers get copied from site to site until the original source is lost or never existed.
How to fact-check a draft
Work through the draft claim by claim. For each one, find the primary source and confirm the claim matches what it actually says, including context and date.
- Highlight every factual claim, number, quote and name
- Trace each one to its primary source, not a site that quotes it
- Check the date: is it still current?
- Check the wording: does the draft overstate the source?
- Confirm product claims with the product team or documentation
- Remove anything that cannot be verified
Common problems
Statistics roundups and AI-generated drafts are frequent sources of errors. AI tools can produce plausible but false figures and citations. Quotes attributed to famous people are often misattributed. Competitor details go out of date quickly.
Fact-check markup
Schema.org defines a ClaimReview type used by fact-checking organizations. Google announced in 2025 that it was phasing out the display of ClaimReview in Search results, so do not expect fact-check markup to produce a rich result there. Fact-checking a marketing article is an internal quality step, not something you need to mark up.
Example
Tracing a popular statistic
A writer at hypothetical company Pebblestrand includes a widely repeated statistic about how much time employees spend searching for information. The fact-checker follows the citation through several blog posts and cannot find the original study. The statistic is cut and replaced with a description of what Pebblestrand's own customer interviews found, attributed accordingly.
Common mistakes
Citing secondary sources
Linking to a blog that quotes a study is not verification. Find the study.
Trusting AI citations
AI tools can invent sources that look real. Open and read every cited source.
Questions
Not as a direct signal. Accurate, well-sourced content supports trust, which Google's guidance on helpful content emphasizes, and avoids errors that damage credibility with readers.
Sources
