Android isn't much different in that regard. I like the concept, and when it's working well, it's a terrific time saver. Sometimes, if I'm undercaffeinated or afflicted with the occasional bout of OMSF (Old Man Shaky Finger), it's awful.
IMHO, it's really awful for the people it's intended to help, ie, those who can't spell to save their life. My wife and sister (not the same person) are two examples. Both are dyslexic, can't spell, and have awful written composition skills. Once they get something wrong more than a couple times, the AI assumes that's what they wanted to write and feeds them the wrong answer forever after.
It's a bigger problem for my sister because I have to help her communicate with doctors, lawyers and bureaucrats regarding my mom's affairs. Too many times, I can't begin to decipher what she's trying to tell me via text. If it's an important issue, I have to resort to a form of 20 questions to work out the definition of some non-word she's thrown at me without context. Often as not, it ends up in a phone call, or even a video chat so she can show me what she's looking at on the computer.
For casual conversations, my sister uses emojis almost exclusively. I have a new-found appreciation for ancient Egyptians.
I've convinced both of them to make more use of screen shots and pictures.
Mrs. Maxx, if I'm out shopping: Get me one of those things we talked about...
Me: What things?
Mrs. Maxx: You know, the thing on TV. They said it was good for weeds.
Me: Who are 'they'?
Mrs. Maxx: Pinterest
Me: I never look at Pinterest
Mrs Maxx, angrily: Oh never mind!
With picture or screenshot + google lens, I've got a reasonable chance of figuring it out.