Fact-checking an interview before publication: a checklist for claims

Checking the recording and checking a claim are separate tasks

Audio can reliably confirm that someone cited a particular figure. It doesn’t show how they calculated it or which period it covers. Check the wording against the recording and its basis against the evidence, especially if the fact affects the article’s central conclusion.

Start with errors that would change the meaning: names and roles, dates, numerical comparisons, the order of events and accounts of other people’s actions. For specialist medical, legal or other complex claims, use a source with the relevant expertise and ask a specialist to review the claim if needed.

Check each claim separately

1. Break statements into parts you can verify

   One sentence can contain a date, a count and a claim about cause. Record them separately. A judgement can also depend on a fact: “best result of the year” calls for checking the metric and period, even if it sounds like an opinion.

2. Find the primary source

   Ask for the document, dataset or published statement that the claim relies on. Prefer a source that directly contains the relevant information. Several retellings of a single report do not give you independent confirmation.

3. Check whether the figures are comparable

   Check event and publication dates, units, time periods, definitions and who or what was included in the sample. More support requests may coincide with having more customers; the numbers alone do not show quality declined. Check causal explanations separately.

4. Record the result and the edit

   Write down the source, the date you checked it and the evidence for the status. If only part of the sentence is confirmed, narrow the claim or state the limitation. If you find a discrepancy, ask the interviewee a specific question rather than correcting the fact based on a guess.

An interview includes a numerical comparison

This is a fictional training example. The names, circumstances and outcomes are included solely to explain the method.

This contains two claims: that the count changed and that a specific cause explains it. To check the first, compare equivalent periods and use the same counting rule. For the second, consider alternatives such as a change in workload or in the types of requests.

Even if the count is confirmed, “the process solved the problem” may go too far. Describe the observed change accurately and present the interviewee’s interpretation separately, with the limits of the data.

Fact-checking log

Give each substantial claim its own row. Another editor can then take over without repeating the research.

Claim and timestamp:
Type: date / number / role / event / cause:
Why it matters to the article:
Primary source:
Source date and event date:
Definitions and units:
What the source directly confirms:
Discrepancies and limitations:
Verification status:
Question for the interviewee:
Final wording:
Checked by and date:

Check the evidence before publishing

• A complex statement has been split into separate claims.
• The source confirms the specific part of the statement at issue.
• Periods and units are comparable.
• Unverified claims are not presented as established facts.