You can avoid this kind of moment by telling the LLM to write in a certain less formal style rather than using its “default voice”.

Personally I'd prefer your own words with GPT tasked to just correct grammar instead of rephrasing it. The GPT-generated rephrasing sounds very off to me due to the formality it introduced in your tone, it flourishes the comment to sound "proper" in a formal tone but completely misses the informal aspect of a comments' section.

I'm not a native speaker either, sometimes I re-read some of my own comments and realise I could have phrased them much better, corrected some grammar but I prefer that they sound like me.

The original text is fine. No need to manicure it via gpt. Who are you trying to please?

It's not who, but what. I just want to be presise. I found the improved comment to be better than my original text, and closer to how I would've formulated myself in my native language.

As I said, there is nothing wrong about using chatgpt, and I use it all the time (also not a native english speaker).

Just found it funny, that even with text is often very easy to tell.