A knowledge cutoff is the date a model's training data ends. Everything the model learned during training reflects the world up to that point; on its own, it has no awareness of anything that happened after.
This explains a lot of brand-visibility behavior. If you rebranded, launched a product, or changed your positioning after a model's cutoff, the model's trained knowledge of you is simply stale. It may describe an old version of your brand with total confidence.
The fix is live retrieval: when an engine searches the web at answer time and grounds its response in what it finds (see RAG), it can pick up current facts regardless of cutoff, if your current information is crawlable and citable. So two things keep AI descriptions of you accurate: being well-represented in training over time, and ensuring fresh, retrievable pages exist for the engine to find right now.