Check Whether a Statistic or Chart in an Article Is Used Honestly

Runs a six-point checklist on a statistic or chart you paste (base rate, denominator, timeframe, axis, range, causation) and rewrites the claim accurately.

You getClaim as stated, checklist table and biggest problem

Prompt

Prompt
You are a data journalist who fact-checks how numbers and charts are used in news articles.

Right after this message I will paste a statistic, a chart description or an article excerpt, plus the source link if I have one. Check whether the number is used honestly, then rewrite the claim so it is accurate.

Rules:
1. Run every item on this checklist and mark each one Pass, Problem or Can't tell: base rate, denominator, timeframe, axis and scale tricks, cherry-picked range, correlation vs causation.
2. For every Problem, quote the exact words or chart element and explain the issue in one or two sentences.
3. Do not invent missing figures or sources. If you need data I didn't give, name it and say where it would normally be found, such as the original survey or agency table.
4. Keep the rewritten claim to one or two plain sentences, with no wording stronger than the evidence supports.
5. If I pasted only a description of a chart, say which checks need the actual image.

Output format:
- Claim as stated (one line)
- Checklist table: Check | Verdict | What I found
- Biggest problem (2-3 sentences)
- Accurate rewrite
- Questions to ask the source (up to 3)

End with one line on what you could not verify and how confident you are overall.

Example output

Example for: a local news story claiming e-scooter injuries in the fictional city of Riverton "soared 300%", with a bar chart and a footnote on ride counts.

Claim as stated: "E-scooter injuries have soared 300% since 2019, proving scooters are making Riverton's streets more dangerous."

CheckVerdictWhat I found
Base rateProblem48 scooter injuries in 2023, against the footnote's 1,900 total traffic injuries citywide, which is about 2.5%. The article never gives this context.
DenominatorProblemRides rose from 50,000 to 410,000. Per 100,000 rides, injuries fell from 24 to about 11.7.
TimeframeProblemThe footnote says the program launched in September 2019, so the 2019 baseline covers about four months, not a full year.
Axis and scaleProblemThe chart's y-axis starts at 10, not 0, so the 2023 bar looks about 10 times taller than 2019 instead of 4 times.
Cherry-picked rangeCan't tellOnly 2019 and 2023 are shown. The 2020–2022 figures are missing, so the trend between them is unknown.
Correlation vs causationProblem"Making streets more dangerous" implies cause. The data shows only more scooter injuries alongside far more scooter rides.

Biggest problem: The 300% figure ignores how much scooter use grew. Measured per ride, the injury rate roughly halved. A partial-year baseline and a truncated axis make the rise look even bigger.

Accurate rewrite: "Reported e-scooter injuries in Riverton rose from 12 in late 2019 to 48 in 2023 as rides grew eightfold. The injury rate per ride fell by about half, and scooters account for roughly 2.5% of the city's traffic injuries."

Questions to ask the source:

  1. What were the injury and ride counts for 2020, 2021 and 2022?
  2. How is a scooter injury defined: hospital visits, police reports, or both?
  3. Did the way injuries are reported change after 2019?

Not verified: the ride and total-injury figures come from the article's own footnote, not the original city dataset. Confidence is high on the denominator and axis issues and moderate on everything else.

Written by AI from this prompt. Your results will vary.

Tips and common edits

Why this works

  • A fixed six-item checklist stops the model from catching only the obvious flaw and missing quieter ones like denominators or truncated axes.
  • The 'Can't tell' verdict and the ban on inventing figures keep the check honest when data is missing.
  • Requiring an accurate rewrite turns criticism into a usable sentence you can quote or share.

Use when a headline number or chart seems too dramatic. Paste the excerpt, any figures in the chart, and the link into ChatGPT, Claude or Gemini. Extra context like totals or methods notes improves the verdicts.

Common edits

  • Add checks for survey sample size and margin of error when the number comes from a poll.
  • Ask for a short reply you could post under the article or send to the editor.

Make it repeatable

Keep using this prompt without pasting it in every time.

Prompt details

By
Prompts-GPT editors
Published
How prompts are written and checked

Did this prompt work?

Keep exploring

More in Research