Short answer
ai product photos marketplace rules are not about whether an image was generated. They are about whether the image misrepresents what a buyer will receive. A generated lifestyle background is usually acceptable. A generated product that does not match the shipped item is a misrepresentation regardless of how it was produced. Reviewers who treat this as an authenticity question will reject the wrong listings and approve the wrong ones.

The distinction that decides most cases
Two questions get conflated in marketplace policy discussions. The first is whether the file is synthetic. The second is whether the image accurately depicts the item. They are independent, and only the second maps to buyer harm.
A studio photograph of a real product on a composited background answers yes to accurate and no to synthetic. A fully generated image of a product that does not exist answers yes to synthetic and no to accurate. A generated image of a real product rendered from its actual measurements can answer yes to both. Sorting by the first question alone will misclassify all three.

Where the real risk sits
- Model and human depiction: generated models wearing real garments raise disclosure questions in some jurisdictions, but the item is still accurate.
- Scale and proportion: a generated image that makes an item look larger, brighter, or more complete than it is crosses from styling into misrepresentation.
- Accessories and contents: bundles rendered to look inclusive when the listing ships fewer items is a common complaint pattern.
- Texture and finish: synthetic renders often smooth surfaces and even out colour in ways buyers later dispute.

What a reviewer should capture
- The listing image as published, archived with its URL and capture time.
- The seller's stated description, so image and text can be compared rather than judged separately.
- Which specific element is claimed to be inaccurate, since "this is AI" is not an actionable finding on its own.
- Whether the seller disclosed generated imagery, if the platform requires it.
Where detection helps and where it misleads
A detector can flag files that do not resemble camera originals, which is useful for sampling a catalogue and prioritising review. It cannot tell you whether the depicted item is accurate, because that requires the physical product and the listing text. A high score is not proof of misrepresentation, and a low score does not establish that the photo depicts the shipped item. Where the detector flags a file but the listing appears consistent, escalate to human review rather than acting on the score alone.
What this workflow cannot do
It cannot verify a product without the product. It cannot recover a listing that was edited before you archived it. And it cannot substitute for the platform's published policy, which differs on whether synthetic imagery must be labelled and on what counts as a material difference. Read the methodology for how evidence layers combine, or check sample reports to see how a mixed-evidence case is documented.
If you need to inspect provenance and metadata on a listing image before deciding, start a free review and record what you find alongside the listing text.