Skip to content

Someone has
to decide.

Before using AI-produced work, check its sources, its fit for the audience, and who could be affected.

Generated curator study
A decision, in miniature · Illustration

The output

It reads beautifully.

An educational draft makes a confident clinical claim. No source has been checked.

Human judgment

Hold the claim.

Fluency is not permission to use it. A qualified reviewer needs to check the source.

Constructed examples, not case reports or automatic tool verdicts.

Four choices. A person’s reasons.

  • Accept

    Fit for purpose as is.

  • Revise

    Usable after named changes.

  • Hold

    Pending missing information.

  • Reject

    Not fit; do not use.

Good
enough?

For whom?

You may see what is wrong without knowing the software that made it.

Before saying yes, ask a qualified reviewer to check the evidence and listen to the people affected.

Give your decision reasons →

Before you
say yes.

  1. 01

    Recognize

    Does it fit the purpose, not just look plausible?

  2. 02

    Interpret

    Separate what you checked from what you assumed.

  3. 03

    Weigh

    Listen to the people affected and consider the stakes.

  4. 04

    Criticize

    Ask a qualified reviewer what misses the mark.

  5. 05

    Test

    Look for what could change your mind.

  6. 06

    Decide

    Accept, revise, hold or reject. Give reasons.

The reviewer is not always the engineer.

An editor, teacher, clinician or creative director may see what an operator misses. Knowing the work matters more than knowing the software when deciding if the result is right.

AI is trained on data. It can help reason, compare criteria and find inconsistencies. That does not decide who can approve the work. A named person answers for the decision.

Fact Interpretation

What a reviewer noticed

The director reported a flaw in a comparison illustration. This is separate from the homepage review. The source image is unverified here.

Illustration

When a near-miss matters

Triage notes, credit decisions, hiring screens, public information.

One free evidence paper

Who gets to say it’s good?

Who can judge the work, how to select fairly, and what AI review can miss.