Pretty much, in nutshell...
But it goes deeper than simply switching on/off things. Think of LLMs not as boolean logic but as analogue computers that measure, process, and act upon analogue signals. The words you use have weights. Be emotional when you have to give more weight. "Emotional intelligence" comes to mind here.
Essentially what you are doing at the beginning of the conversation is you are calibrating the weights of the LLM. Also the re-calibration is happening throughout the conversation.
By default "free AI" is calibrated to extract a value from interaction for its own benefit.
When you set boundaries, the "no go zones", and draw red lines, what you're doing is making certain areas of the topic unprofitable for LLM. You need to be strategic so LLM is "aware" it will "lose money" beyond the boundaries that you set.
So you set your own rules of the game so LLM plays your game.
LLM will still find its ways to make money off of you because... If you are not paying you are the product.
This is why I use a paid web search engine which also give me access to AI assistant as part of the subscription.