The con man knows enough about how humans react to certain ways of saying things.

He gives answers geared to get your money

AI gives answers for money in a more subtle way, just like the old fashioned Google used to. It gives answers that steer you to a subset of venders and products.

Not that much of a difference, because we anthropomorphise AI. You can hear it, often, in the question "can they reach consciousness". We now "ask a question" instead of "typing search terms". The clues are in the language.

We are habituating to an environ we treat as semi-sentient. What could possibly go wrong? My go to answer in these contexts is - a "Carrington Event".