Why it definitely IS possible that Google listens and uses what we say for ads
When reading about coincidences where people got ads of things they talked about recently, I often read comments like "It is just not worth it for Google to transmit that mass amount of audio data to its servers".
That is surely true, but Google doesn't even need to do that. Google Assistant for example already comes with a built in language model and listens for a "Hey Google" all the time anyways. All it needs to do is to extract the essence of what we say, which is summarizable into few or single words like "BMW" or "Carrot", and send it to Google's servers. Suddenly, the mass amount of data is decreased to a few kb over the day. They don't need everything, tags like "Climbing", "Motorbike" or "Ill Cat" are already enough to build a profile and follow your current situation.