Which images in a mixed real-and-AI product library can support a factual claim?

Separate image origin from evidentiary authority

A mixed product library may contain original photographs, edited composites, generated scenes, and files with unknown origins. Knowing which file produced another explains its history, but does not establish what the resulting image can prove. The editorial decision is whether an asset supports a specific factual statement.

Consider a fictional kettle library containing an original label photograph, a background composite, and an AI-generated internal view. The original photograph may support reading the visible label if the product identity and text are independently checked. The composite may present the kettle in a different setting. The generated interior cannot establish the kettle's actual construction.

Do not promote a convincing appearance into evidence. A realistic image may still contain invented or altered product details.

Assign an evidence class and permitted use

Use four operational classes: measured or directly documented capture, unchanged-product composite, generated illustration, and unknown provenance. These are editorial classifications, not technical certifications. Assign permission to individual claims rather than granting blanket approval to an entire image.

A directly captured label can support inspection of visible wording, but not an unseen heating mechanism. A composite retains limited evidentiary value only where the relevant product details can be checked against the source. Unknown provenance means the reviewer cannot establish enough about the asset to rely on it.

Test each proposed claim against the asset

A single image may support one observation while failing to support another. For the fictional kettle, the original label photograph could establish what text appears on that photographed unit. It cannot independently confirm that every production unit has the same label or that a printed performance statement is true.

Likewise, a background composite should not become the source for material, dimensions, or hidden components merely because the visible kettle looks unchanged. Keep the factual reference separate from the presentation file.

Itemlume produces raster visual concepts. Its generated images do not establish physical product properties. Text planning is free after sign-in; image and video generation follow current quoted prices.

1. Write the exact factual claim being considered.
2. Identify which visible feature or record could support it.
3. Check whether that information came from a direct capture, preserved source, or generation.
4. Compare relevant composite details manually with the original.
5. Approve the specific evidentiary use, restrict it, or request another source.

Maintain a claim-specific release register

Record the asset identifier, evidence class, permitted claim, restriction, reviewer, and decision date. Do not substitute a generic 'approved image' label for these distinctions. A photograph approved for label transcription is not automatically approved as proof of performance.

Before publishing, inspect the final crop, caption, and surrounding copy. A cropped label may no longer show the detail that justified its evidentiary use. If an image is edited again, reconsider whether the relevant information remains intact.

Shopify product media can deliver visual content, but publication does not verify factual accuracy. Evidence classification and approval remain manual editorial responsibilities. Keep unsupported claims out of the final creative until suitable evidence is available.

Record the decision in a manual worksheet

Manual Product Image Evidence Classification Register

The download contains fictional practice rows. Actual evidence and release decisions are blank for you to fill in. This manual record does not inspect products, update a store or certify compliance. Keep private material out of public copies.

Download the English CSV worksheet