Skip to main content
Lesson 10 of 12 · Deep Research and Web Search

Fact-checking: verifying information

Lesson 10 of 12 Updated November 5, 2026 Intermediate
On this page
  1. Verify quickly, verify well
  2. The SIFT framework
  3. Tracing data back to its source
  4. Rely on professional fact-checkers
  5. Your turn
  6. Key takeaways

Verify quickly, verify well

Fake studies, invented figures, truncated quotes, and images taken out of context: misinformation spreads faster than ever. You face it daily, often from well-meaning people who share without checking. Everyone knows they should verify; what’s missing is a method fast enough to actually use—two minutes, not an hour. ChatGPT with web search makes this timeline achievable: it finds the original source, checks what other publications say about it, and presents the results clearly. The key is to run your queries in the right order. Otherwise, you’ll mostly just confirm what you already thought.

The SIFT framework

The SIFT method, developed by Mike Caulfield, consists of four moves. Stop: pause before sharing or believing a claim—this is the hardest step, since sharing almost always precedes thinking. Investigate the source: who is saying this, and is that source reliable? Find better coverage: do other trustworthy sources report the same thing? Trace claims: go back to the original source of the assertion. ChatGPT can assist you at each of these stages, provided you keep them distinct instead of asking for everything in one go.

Wording matters. A leading question biases the search. Ask, “Is it true that [claim]? Find reliable sources that confirm or refute this information,” not “Prove that…”. The neutrality of your question determines the neutrality of the answer. This applies to routine checks like “Is it true that France was the world’s top wine producer in 2026?” or “Did [public figure] actually say [quote]?”

Tracing data back to its source

A circulating statistic almost always has an origin, and that origin often says something different from what people claim. Frame your query like this: “The figure stating [statistic] is frequently cited. What is the original source of this number? Is it still current?” The second part matters as much as the first: a percentage that was accurate in 2020 may be false today, yet articles keep repeating it without updating.

The same approach works for quotes, where false attribution is an old and widespread problem. Asking “Did [public figure] really say [quote]? What is the original source?” eliminates many fanciful attributions. For an image or video that seems suspicious, the most effective phrasing leans on professional fact-checkers: “Is this image circulating on social media showing [description] authentic? Have fact-checkers flagged it as false or misleading?”

Rely on professional fact-checkers

Several newsrooms dedicate themselves exclusively to this work, and naming the right ones in your prompt significantly improves response quality. AFP Factuel is the verification service of Agence France-Presse; Les Décodeurs at Le Monde verify information in French; Snopes is an international reference; Full Fact conducts rigorous work in the UK; PolitiFact focuses on political statements in the English-speaking world. A query like “What do fact-checkers (AFP Factuel, Les Décodeurs) say about [claim]?” gives you access to investigations already completed, often better documented than what you could produce alone.

The SIFT method Four steps: Stop, pause before believing or sharing; Investigate, examine the source; Find, locate better coverage of the topic; Trace, go back to the origin of the claim. FACT-CHECKING The SIFT method S Stop Before sharing I Investigate Who is speaking? F Find A better source T Trace Back to the origin Phrase the check neutrally: "Is it true that…", never "Prove that…".
Mike Caulfield's SIFT method, applied to research with ChatGPT.

Your turn

Put SIFT to the test. Take a piece of information you recently saw on social media or heard in conversation, and run through SIFT without skipping a step: don’t share it yet, ask who the source is, search for confirmation from other reliable outlets, then trace it back to the primary source. Compare the result with your initial gut feeling—when they differ, that gap is the best lesson this course can offer.

Three pitfalls lie in wait. The first is confirmation bias, which leads you to seek only what supports your existing opinion: explicitly ask for counterarguments. The second is the volume fallacy: a falsehood repeated a thousand times remains false. The third is ignoring context: an accurate statistic can mislead if you omit what it measures and over what scope, which is why you should always demand full context. Finally, remember that not everything is verifiable—some claims are opinions, not facts. Knowing the difference is part of the job.

Official sources: Does ChatGPT tell the truth? · Searching the web with ChatGPT (OpenAI Help Center)

Key takeaways

  • Apply the SIFT method: Stop, Investigate, Find, Trace
  • Phrase your checks neutrally to avoid biasing the search
  • Always trace a statistic or quote back to its original source
  • Use recognized fact-checking sites as references
  • Treat fact-checking as a habit to build, not a one-off task