Get AI to catch its own mistakes
AI is far better at reviewing its own work than getting it perfect first time. Learn the two-pass move.
What you’ll do
- Take a draft AI just produced.
- Ask it to find the three weakest things about it.
- Have it fix them and explain what changed.
Common questions
Does asking AI to check itself actually work?
Yes — it catches a lot, especially vague wording, missing steps, and overconfident claims. It won't catch everything, but a two-pass approach produces noticeably cleaner drafts than a single pass.
Doesn't this mean AI is checking its own homework?
It is, which is why you still read the result yourself. Think of it as a quick self-edit before it reaches you — it removes the obvious issues so your review focuses on the things that really need your judgment.
What if the AI says its draft is fine but it isn't?
That can happen, especially with facts or figures that need outside verification. The lesson covers what to always double-check yourself — anything numerical, anything you'll put your name on.
Will this stop AI from making things up?
It reduces it. Asking AI to flag anything it's uncertain about catches some invented details, but it won't catch all of them. For anything factual, you still need to verify from a reliable source.
Is there a risk it makes the draft worse when it 'fixes' things?
Occasionally. The lesson shows how to ask it to explain what it changed, so you can accept or reject each fix rather than just taking the revised version blindly.
How long does the two-pass approach add to the process?
Usually under a minute — you're adding one follow-up prompt. The time saving on your own editing more than makes up for it.