> What is most likely to happen is the so called Answer Engines will embed advertising into their results -- except in a more insidious, subtle, hard to detect and filter out manner.

This is the part of the new AI paradigm that concerns me the most, and I think you are correct.

If "pay for placement" (or worse "legislate for placement") in LLM training becomes a thing, then we lose all transparency as these biases get baked into the knowledge set, and users have no way of knowing where and when they get applied.