Side by side
| | PIM | Canonical product record | | --- | --- | --- | | Where it lives | inside the company | on the public web, at a stable URL | | Who reads it | your teams and your channels | AI systems, crawlers, anyone checking | | What it holds | every attribute you manage | identity and the facts a buyer asks about | | Provenance | internal edit history | the source page, the date read, a digest | | Conflicts | resolved internally, one value survives | kept visible: who said what, when | | Output | feeds to retailers and marketplaces | a record other copies can point back to |
Why one does not replace the other
A PIM is where the truth is decided. It is not where a model can see it: the model sees what the channels published, in whatever shape each channel chose. A canonical record is the published, checkable form of what the PIM decided — with the evidence attached, so that when a retailer's listing disagrees, a model has something better than a majority vote.
When you need both
When your products are described on many sites you do not control — which is to say, when you sell through retailers. In answers about real products, 37% of citations pointed at the maker's or seller's own site (study); the rest is where the PIM's decisions arrive second-hand. What a record may and may not be used to claim is set out at /evidence.
Sources
- Canonical product record
- Whose pages AI cites when it answers about a product
- Evidence methodology, version 1.0