When does using AI stop being collaboration and start being outsourcing your thinking?

I've been thinking a lot lately about how I personally use AI tools in creative and professional work, and I'm genuinely curious where others draw the line. There's a real difference between using AI to brainstorm, get unstuck, or pressuretest ideas versus just outsourcing the thinking entirely. But in practice that line gets blurry fast. You start by asking for a rough outline, then you're editing its draft, then you're mostly just accepting suggestions, and somewhere in there you've lost the thread of your own voice. What I find interesting is that the most satisfying results I've had come when I treat the AI like a sparring partner rather than a ghostwriter. Push back on it, argue with it, use its output as a foil. That seems to produce something that still feels like mine. But I also wonder if that's just me rationalizing a comfort zone. Maybe the resistance to full AI collaboration is just ego, or maybe it's something worth protecting. For people who use AI regularly in creative or knowledge work, how do you think about this? Do you have explicit rules for yourself, or does it just depend on the task? Has your approach shifted over time as the models have gotten better? Curious whether others feel like the better the AI gets, the harder this question becomes.