If a user clicks on one ad, OpenAI makes a tiny bit of money. If a user clicks on 10 ads, OpenAI makes 10x the money. This incentive is inevitably going to lead to ad supported LLM's not answering questions properly the first time so more ads served. I am guessing in voice mode they will also have voice ads.
If I can tell that the LLM didn't answer my question properly, why would I engage with the ad the answer contains, which would likely be equally irrelevant?
They have behavioural experts who know how mainstream minds work and how to manipulate.
Remembers ads are predominately not for the educated and intelligent, that type of person will think critically. People who are more naive are the golden gooses