Unlike Search, chatbot advertising can use advertiser-supplied audience data and conversion signals to surface products unrelated to the content of non-commercially motivated product usage.

This is a *choice* that is made by Search. There's no reason that Google couldn't show ads for non-commercial queries that are based on prior searches, or externally-sourced advertiser-supplied information.

The reason they don't – at least based on what they're said publicly – is that non-query-relevant searches are statistically detrimental to repeat usage of their tools. Over time, they've modeled that non-relevant searches results in fewer sessions overall.

Is there a reason that this dynamic would NOT hold for chat interfaces?

That's fair. I suppose I could have worded that to say that using those signals is likely workable (meaning: doesn't erode or inhibit long-term retention), given the product experience.

As to why that's the case: the use cases are fundamentally different. Search is successful when it quickly ingests a short query and directs the user to the most appropriate destination. A chatbot is designed for conversational interaction across potentially disparate contexts in a given session; it is the destination. So my sense is that the short intended interaction time in Search would make ads that are not immediately relevant to the context more distracting than in a chatbot.