Content Credentials work like a tamper-evident label inside the file. Each tool that supports the standard adds an entry, such as captured by this camera, edited in this program or generated by this model, and signs it. A viewer that supports the standard can show the history and tell whether the record has been altered since.
The aim is provenance: a way to check the origin of a picture that does not depend on how it looks. The record differs from a visible watermark, which sits on the image, and from a plain metadata field, which anyone can rewrite without a trace.
Support is still uneven. Not every camera, editor or platform writes or keeps the record, and an image without Content Credentials is not thereby suspect. For AI-generated product images, the marking that search engines and marketplaces name most often is metadata such as the IPTC source type field; Content Credentials are a second, stronger layer where the tools in use support them.