Knowledge cutoff

Every LLM has a knowledge cutoff: the point up to which its training data extends. Anything your brand did after that date — a rename, a launch, new pricing — is unknown to the model in trained-data mode. This is why young companies and recently-rebranded ones often score poorly on trained-data audits even when their live-search results look fine.

You can't change a cutoff, but you can compensate: maximize retrievability for web-search mode now, and build the durable footprint (Wikipedia-grade references, consistent entity data, wide coverage) that the *next* training runs will absorb.