Skip to content

Reviewing AI-generated images

Images generated by a model do not go straight into your store. They wait. The screen says why in one sentence: “Images a model drew are unusable until you have looked at them. Check the product before approving.”

Until you approve one it carries the badge Awaiting review, and in the media library it is labeled AI — awaiting review rather than simply AI-generated.

The gate is deliberate, and it is not about picture quality

Section titled “The gate is deliberate, and it is not about picture quality”

A generated image can be beautiful and still be wrong. The instruction on the screen is specific: check the product, not the picture. Models produce plausible objects — a jar with the wrong number of clasps, a label whose text is nearly your brand, a fabric weave that is not the one you sell.

A shopper does not experience that as a rendering artifact. They experience it as a product that did not arrive as shown, and that is a return, a refund and, in most markets, a misdescription.

Nobody but you knows what your product looks like. That is the whole reason the approval step exists and why it cannot be automated away.

One message on this screen is worth reading twice:

“This file has no machine-readable AI marking. Generate it again before publishing.”

A generated image is supposed to carry a marking that identifies it as machine-made — readable by software, not just by a human looking at a caption. If it is missing, CIQRA tells you to generate the image again rather than letting you approve it anyway.

That is a compliance posture as much as a technical one: rules on labeling AI-generated content are arriving, and a store that quietly published unmarked generated imagery has no way to find those files later. Regenerating now costs a minute; identifying them afterwards means looking at every image you ever published.

“Nothing to review — generated images appear here before they can be used.” If the queue is empty, nothing is waiting on you. It is not a screen to check daily; it is a screen that fills when you have been generating.

  • It does not cover generating the images. This screen is the gate between generating and using, not the tool that makes them.
  • It does not say what the marking is, technically, or which readers understand it. That belongs with the generation feature and with the standards it follows, both of which move.
  • It does not tell you your legal obligations for labeling AI content where you sell. They differ by market and they are changing; that is a question for your own counsel rather than for us.
  • It does not say whether an approved image can be un-approved, or what happens to a rejected one. Neither has been checked against the current screen.