Do not rewrite everything at once

When an image is close but wrong, beginners often throw away the prompt and start over. That makes iteration random. Professionals diagnose the problem and change one thing at a time.

Think in categories:

• Subject problem: the wrong thing is visible.
• Style problem: the image feels off-brand.
• Composition problem: the crop or focal point is wrong.
• Accuracy problem: the diagram or scene teaches something false.
• Artifact problem: text, hands, UI, or details are distorted.

Targeted revision examples

Instead of:

Make it better.

Use:

Keep the same composition, but simplify the background and remove all readable text.

Instead of:

Try again.

Use:

Regenerate in the same monochrome editorial style, but make the workflow cards larger and reduce the number of objects on the desk.

The acceptance checklist

Accept an educational image only if:

• It supports the lesson objective.
• It is visually consistent with the course.
• It has no misleading technical details.
• It works at the intended crop size.
• It has clear alt text.
• It does not rely on fake logos, fake UI, or fake claims.

Image generation brief

Asset: Art director review checklist Prompt: Create a clean editorial illustration of an art director reviewing three AI-generated course images on a large monitor, with checkmarks beside composition, clarity, consistency, and accuracy. Monochrome educational style with tonal contrast for emphasis, calm workspace, no readable paragraphs, no logos, no watermark. Alt text: A reviewer checks AI-generated course images for clarity, consistency, composition, and accuracy.

Try this: Pick one image you almost like. Write only one revision instruction. If you change five things at once, you will not know what helped.