Even discounting this, and despite everyone bleating on about its (very real) flaws, ChatGPT and other LLMs do quite a good job of proofreading and suggesting improvements to written English text[0]. I find it works best if you keep them on quite a tight leash but it's certainly within the compass of their capabilities to take badly written English and turn it into well written English, and even adopting a particular style to do so.

[0] Performance in other languages... well, I suspect it's still going to be quite variable, which is another valid criticism that has been levelled at the more popular mainstream models over the past year or two.