Skip to main content
AI Safety, Alignment & Ethics

⏱ About 15 min15 XP

Fact-Check Challenge

You have spent this module learning the theory: how AI hallucinates, how deepfakes work, how misinformation spreads, how to check sources, and how to think skeptically. Now it is time to put all of that into practice. This lesson is built around a hands-on challenge. You will work through a series of claims — some true, some false, some misleading — using the verification toolkit you have built. This is the kind of work that professional fact-checkers do every day.

Your Fact-Checker's Toolkit Recap

Before you begin, here is a rapid summary of the tools at your disposal. Lateral reading: leave the source and search for what independent outlets say about the claim and publisher. Reverse image search: drop a suspicious photo into Google Images or TinEye to see where it has appeared before. Primary source tracing: if a claim cites a study or expert, find the original document or statement directly. Fact-checking sites: Snopes, PolitiFact, FactCheck.org, Reuters Fact Check — search these before doing any other research. The pause questions: Is this actually true? Who benefits if I believe it? Have I verified it? Would I want this attributed to me? AI hallucination awareness: treat any specific citation, statistic, or name provided by an AI as unverified until independently confirmed.

Fact-Checking Order of Operations

Start with a quick Snopes or FactCheck.org search — if the claim has been widely circulated, it may already be investigated. Then trace to primary sources. Use reverse image search for visual claims. Lateral reading is your fallback for anything not yet covered by dedicated fact-checkers.

Challenge Set A: Evaluate the Claims

The following four claims represent the kinds of things that circulate online. For each, you will apply your toolkit in the main activity below. Read all four now before you begin investigating. Claim 1: 'Scientists have discovered that drinking two glasses of water immediately after waking up cures insomnia, according to a Harvard Medical School study published last year.' Claim 2: 'A photograph shows a massive crowd gathered in a city square during a protest last week against a government policy.' Claim 3: 'An AI assistant, when asked about the history of jazz music, stated that the saxophone was invented in 1820 by Johann Schmidt in Berlin.' Claim 4: 'Studies show that students who take handwritten notes remember information significantly better than those who type notes on a laptop.'

Why These Four Claims?

These claims represent four different verification challenges: a specific medical claim with a named institution, a visual claim about an event, an AI hallucination in a historical domain, and a research-backed educational claim. Each requires different tools from your toolkit.

For Claim 1 — the Harvard study on water and insomnia — what is the FIRST verification step you should take?

For Claim 3 — the AI stating the saxophone was invented by Johann Schmidt in 1820 — what type of AI error does this most likely represent?

The Full Fact-Check Challenge

  1. This activity asks you to work through all four claims from Challenge Set A using your complete verification toolkit. Budget about 15-20 minutes.
  2. For EACH of the four claims, complete the following steps and record your findings:
  3. Step 1 — Claim type: Identify what type of claim this is (medical/scientific, visual/photographic, AI-generated historical, or research-backed). This helps you choose the right tools.
  4. Step 2 — Quick fact-check search: Search the claim on Snopes, FactCheck.org, or PolitiFact. Write down what you found (covered / not covered).
  5. Step 3 — Primary source hunt: For Claim 1, search for the Harvard study. For Claim 2, do a reverse image search on a similar image or describe what you would do. For Claim 3, look up who actually invented the saxophone and when. For Claim 4, search for research on handwriting vs. typing notes.
  6. Step 4 — Lateral reading: For any claim not already verified, search for what three independent sources say about it.
  7. Step 5 — Verdict: Rate each claim as True, False, Misleading, or Unverifiable. Write one to two sentences of evidence for your verdict.
  8. Step 6 — Reflection: After completing all four, answer these questions: Which claim was hardest to verify and why? Which tool from your toolkit was most useful overall? What would have happened if you had shared all four claims without checking?
  9. Bonus: Create a fifth claim of your own — one that is intentionally misleading but not completely false — and swap it with a classmate to fact-check.