Skip to main content
The Human BitAsk Bit

Lesson 05 · 10 minutes

Check before you trust

Match the checking method to the consequence: inspect sources, calculations, assumptions, omissions and edge cases.

Design a proportionate review instead of accepting polished output.

Why this matters

Generated answers can combine true details, weak inference and invented information in one fluent response.

Review cost determines whether an apparent time-saving remains worthwhile.

The practical bit

Check facts against an authoritative source, recalculate important numbers, compare the answer with the original material and inspect what may have been omitted. Ask for sources only as a discovery aid, then open and verify them yourself.

Create a short checklist for recurring work. If reliable checking takes longer than doing the task conventionally, change the workflow, reduce the scope or do not use AI there.

What it looks like

At work

For a policy summary, compare every material obligation and date against the actual policy. Do not review only whether the prose sounds professional.

In everyday life

For a product comparison, verify current price, warranty and compatibility on primary sources before spending money.

Common mistake

Asking the same model ‘are you sure?’ and treating its second confident answer as independent verification.

Use your judgement

What evidence would let you detect a plausible but wrong answer?

Try it now

Build a three-point check
  1. Choose one generated output.
  2. Identify the three errors that would matter most.
  3. Name an independent source or test for each error.
Check: Your checks should test the work, not merely ask the generator to endorse itself.

Use this in a real situation

Review a research briefingTrace claims, inspect sources and decide what is safe to reuse.Open the practical →

Keep this

The checking method is part of the workflow and part of the cost.