AI Guides › Prompting
The Habit Of Asking It To Grade Its Own Answer Before You Do
By Nigel Guy · 1 min read
Most people read an answer once and decide, immediately, whether it's good.
That first read is coloured by relief at having an answer at all, which
makes it a worse judge of quality than a second, more deliberate pass —
including one where you make the model do the deliberate pass first.
The rule: before you judge an answer yourself, ask it to critique its own
draft against the original brief — a self-critique surfaces gaps a first
read tends to miss.
The mechanism
- Get the first draft, as normal.
- Ask directly: "review this against what I originally asked for — what's
missing or weak?" Not "is this good?" — a specific instruction to
compare the draft against the actual brief.
- Read the self-critique before you read the draft again. It reframes
what to look for on your own second pass.
- Decide which flagged gaps actually matter, and which are minor — the
self-critique surfaces candidates, it doesn't set priority for you.
- Only then do your own review. The self-critique is a first pass, not
a replacement for your own judgement.
What to skip
Skip trusting the self-critique as the final word — a model reviewing its
own work shares the same blind spots that produced the original draft, so
it won't catch everything. And skip doing this for trivial, low-stakes
outputs where a quick read is genuinely sufficient.
Guardrails
- This catches gaps against the stated brief. It won't catch a wrong
underlying fact or a flawed premise the brief itself didn't ask about —
factual verification is still a separate step.
- A self-critique that finds nothing wrong isn't strong evidence of
correctness — treat "no issues flagged" as a weak signal, not a clean
bill of health.
- This is one useful check among several, not a substitute for your own
final review, especially on anything that matters.
All 751 AI guides · JulieMango plans from £17/mo