HUNGRYAI · PRACTICAL LEARNING GUIDE

Before you trust an AI answer, check these things

A confident answer can still be incomplete or wrong. Treat AI output as material to review. A short, deliberate check can help you decide what is useful, what needs revision, and what you should not rely on.

Does it answer the actual question?

Compare the output with your original task. Did it keep your constraints? Did it address the intended audience? A fluent response is not useful if it solves a different problem.

Can you verify the claims?

Identify claims that matter to the outcome. Open the original sources when sources are provided, and check that they support the claim. Look for relevant dates and context. A citation that looks convincing is not proof that the source exists or says what the answer suggests.

What assumptions are hiding in the answer?

Notice where the response fills in missing details. Does it assume a location, budget, skill level, or goal you did not provide? Make those assumptions explicit before you use the result.

Do the details hold together?

Recalculate important numbers independently. Check names, dates, units, and totals. For an instruction or plan, walk through the steps and look for missing dependencies.

That second response also needs review. Asking the same tool to critique itself does not replace independent verification.

Should a person make the final call?

Consider who could be affected if the answer is wrong. When a result shapes an important decision, involve someone with the relevant expertise. Remove information that should not be shared, and keep the final decision with the responsible person.

A small checklist to keep

• Matches the task and audience.
• Important claims checked against reliable sources.
• Assumptions and missing information made visible.
• Numbers and practical steps reviewed.
• Appropriate human review before use.

Next: write a clearer prompt →