I strongly doubt that voice assistants are sitting unused because people are instinctively resisting profiling. People don't seem to be instintively resisting profiling when using Google, or Facebook, or indeed anything else.
Rather, I think the problem is simply that they just don't have that many uses. Most of the use cases that their developers thought out can be accomplished both faster and more reliably with a screen and keyboard.
Look at what the two common voice assistant tasks are. "Play music" and "Set a timer" are requests that are useful when I'm busy doing something else away from my device. I.e., they are tasks not worthy of my primary attention. Also, they are both tasks where if the voice assistant screws up, I can immediately tell (when I hear the first few notes, or when it reads back the timer wrong) and I've lost nothing - not even the five seconds enunciating the query, because, as mentioned, I was doing something else anyway. There just aren't that many tasks with that profile.
Why would I ask Assistant to search Google for something, when I can get the results faster by typing the query AND without wasting time repeating it three times until the voice recognition gets it right?
Why would I ask Alexa to buy something on Amazon, when I can search Amazon faster on the website AND without risking ordering the wrong thing by mistake?
I bet that the reason Apple didn't actually make the Knowledge Navigator, and the reason they didn't bet the farm on Siri, is that if there's one thing Apple is good at, it's UX engineering. They understand that if you give users a really cool way to do in ten seconds something that can be done in five seconds in the boring old way, they'll switch back to the boring old way in an hour.
What voice assistants need in order to actually get used is to be like a real live PA. Someone who, when asked to Google something, will not just Google what I said and read the first hit, but also know from previous context what I'm really looking for, then read through the query results, refine the query and exclude irrelevant hits, collate the information and present it to me in a synthetic fashion. Someone who, when asked to buy a gizmo, will not just search Amazon and order the first hit for 'gizmo', but know the specific details I want, search results to find a gizmo that matches them, do price-comparison, read reviews and figure out which ones are reliable, and pick the most convenient shipping option based on my schedule.
ChatGPT looks like it could do that at first glance... but it can't. First of all, a model like ChatGPT is trained once and does not learn further after training. Sure, it can tell you about WW2, but go ask it about something that happened the day before yesterday; it won't know. And retraining is computationally extremely expensive.
Secondly, ChatGPT always sounds convincing, but is actually wrong a lot. Fine for a laugh, but I wouldn't trust it to do anything that I actually want to get right.
In short, voice assistants are a no-starter until we actually have strong AI. I'm not holding my breath.