Citations and source trails

A citation is a claim about evidence, not evidence by itself. It becomes useful only when the source exists, supports the statement, and fits the context.

Use this three-part method:

1. Open every important source. Write down what this means in the scenario, what evidence would show it was done, and who owns the decision.
2. Match the quoted claim to the source passage. Write down what this means in the scenario, what evidence would show it was done, and who owns the decision.
3. Check date, author, scope, and source quality. Write down what this means in the scenario, what evidence would show it was done, and who owns the decision.

Walk through An essay draft includes three realistic journal titles; one exists, one has the wrong author, and one was invented.. Label three moments where “Citations and source trails” changes what you trust: (1) the first fluent answer, (2) the first missing source or permission, and (3) the decision a human must own. Write the before/after task so the model only does the slice that evidence supports. Keep one sentence that states how this page’s idea differs from a generic “AI is smart/dumb” score.

• Counting links instead of reading them. This shortcut removes useful friction, but it also hides an assumption that should be tested. Replace it with an observable check.
• Accepting a source that discusses the topic but not the claim. This shortcut removes useful friction, but it also hides an assumption that should be tested. Replace it with an observable check.
• Using an ai-generated quotation without locating the original. This shortcut removes useful friction, but it also hides an assumption that should be tested. Replace it with an observable check.

Before you close the tab, capture a reusable habit for Citations and source trails inside Why AI makes mistakes: name the observable check, the evidence you would open, and the stop condition. Rehearse it once on a low-stakes example, then once on a higher-stakes variant. The habit succeeds when you can explain the check without reopening this lesson. Target outcome: Explain why plausible language generation does not guarantee factual accuracy.