AI Prompts for Fact Checking
Here are strong AI prompts for fact-checking that you can copy and reuse.
1. General fact-checking prompt
Fact-check the following claim carefully.
Claim: [INSERT CLAIM]
For each factual statement:
1. Determine whether it is True, False, Misleading, Partly True, or Unverifiable.
2. Explain why.
3. Identify what evidence would confirm or disprove it.
4. Prefer primary sources, official records, peer-reviewed research, government data, and reputable news organizations.
5. Distinguish established facts from interpretations, opinions, and predictions.
6. Flag outdated information or claims that depend on a specific date.
7. Do not assume the claim is true simply because it sounds plausible.
Give me a concise fact-check with sources and a confidence level for each conclusion.
2. Prompt for checking an article
Fact-check this entire article.
[PASTE ARTICLE]
Extract the article’s significant factual claims and evaluate each one independently.
Use this format:
– Claim:
– Verdict: True / Mostly True / Partly True / Misleading / False / Unverifiable
– Evidence:
– Source:
– Context or missing information:
– Confidence:
Pay particular attention to statistics, dates, quotations, causation claims, scientific claims, and statements attributed to people or organizations.
Do not fact-check opinions unless they contain an implicit factual assertion.
3. Check claims against primary sources
Act as a rigorous investigative fact-checker.
For the claim below, prioritize primary sources over summaries or secondary reporting.
Claim: [INSERT CLAIM]
Search for:
– Government documents and official statistics
– Original research papers
– Court records
– Corporate filings
– Official statements
– Original datasets
– Direct transcripts or recordings
Then:
1. State exactly what the primary evidence establishes.
2. Identify what it does NOT establish.
3. Compare the evidence with the original claim.
4. Give a final verdict.
5. Provide links to the most authoritative sources.
4. Detect misleading statistics
4. Analyze this statistic or numerical claim for accuracy and misleading presentation.
Claim: [INSERT STATISTIC]
Check:
– Original source
– Sample size
– Time period
– Geographic scope
– Denominator/base rate
– Absolute vs. relative change
– Selection bias
– Missing context
– Whether correlation is being presented as causation
– Whether the comparison is appropriate
– Whether the statistic has been updated or superseded
Calculate the relevant numbers yourself where possible and explain the result in plain English.
5. Fact-check a social media post
Fact-check this social media post as if you were preparing a professional fact-check.
[PASTE POST]
Break it into individual factual claims. For each claim:
– Quote or summarize the claim
– Give a verdict
– Explain the evidence
– Identify the strongest available source
– Note any important context
– Identify misleading wording, cherry-picking, or omitted information
Separate factual errors from reasonable disagreements or opinions.
Do not exaggerate the conclusion beyond what the evidence supports.
6. “Steelman before you fact-check” prompt
Fact-check this claim fairly and without assuming either side is correct.
Claim: [INSERT CLAIM]
First, state the strongest reasonable interpretation of the claim.
Then identify:
1. Evidence supporting the claim
2. Evidence contradicting the claim
3. Important uncertainty
4. Definitions or assumptions that could change the verdict
5. Whether the available evidence is sufficient for a definitive conclusion
Give the most nuanced conclusion supported by the evidence rather than forcing a True/False answer.
Best practice: ask the AI to show its sources and distinguish “I couldn’t verify this” from “this is false.” Those are very different conclusions.
