Foundation model
A foundation model is a large AI model trained on a huge amount of general data, which other products and assistants are then built on.
A foundation model is a large AI model trained on a huge amount of general text, and often images and other data, that can then be used for many different jobs. The assistants people ask about your brand, like ChatGPT and Gemini, are built on foundation models.
Knowing the term helps you see why the same question can get different answers in different places. Different products may sit on different models, and the same model can behave differently depending on what's built around it.
Model and product
The model is the part that writes text. The product is everything around it: the app, the web search, the rules about what to show, the memory of past chats.
For example, two assistants could use the same foundation model, and one searches the web before it answers while the other answers from memory. Asked which roaster has the best decaf, the first might cite a review from last month and name Kestrel Roasters. The second might name a brand that was well known when the model was trained.
So when you track answers, you're tracking the product, with whatever model it picked that day.
Why they change
Foundation models are expensive to make, so there are only a handful of companies making the biggest ones. They release new versions every so often, and the products switch over. A new model can describe your category differently, even if nothing about your brand has changed.