And yet, Bard seems to be worse than ChatGPT. Google aren't the only ones who can crawl the web. Given the depth of their index and how much spam it contains it might not even be the asset it seems.

In particular because they are more or less using a term index I believe. This new world relies on vector indexes for semantic search.

I'm not sure the quantity of data will be a differing factor once the LLM reaches a certain point of parameters.

In this context I think we’re talking about moat, i.e., private data, that they can leverage for personalized experiences. Similar to what Microsoft announced with their Office365 Copilot stuff.