Marketplaces and platforms increasingly check how an image was produced, and the EU AI Act requires generated media to say so in a way a machine can read. Until now an image you downloaded from Qamera said nothing about its origin beyond what the model provider had put there.
Every image generated from now on — packshots, photo shoots, model cards and sceneries — carries an IPTC DigitalSourceType declaration inside the file, recording that it was produced by a trained model. Images that depict a synthetic person additionally carry the keyword Amazon requires for that case; product-only images do not, because a declaration that is not true of the picture is worse than none.
The declaration is written into the file without re-encoding it. Your image is byte-for-byte what the model produced, apart from the small metadata block that was added — the resolution, the compression and every pixel are untouched, and the provenance information the model provider embedded stays intact alongside ours.
Two things worth knowing. The declaration travels with the original you export, not with the smaller preview sizes shown inside the application. And once an image is uploaded to a shop, whether it survives is up to that shop: some e-commerce platforms strip all metadata from product images when they re-process them, and nothing embedded in the file can prevent that.
Images generated before this release are unchanged. Photos you upload yourself never receive an AI declaration.