The more colloquial, sloppy, and nonstandard the language, the less likely a student's cheating will be flagged as plagiarism. There are many, many dishonest uses for LLMs; students around the globe are leveraging their ability to furnish essay answers and entire reports with expert-level domain knowledge.
If a student doesn't engineer the prompt properly and the LLM uses its "native" voice, this can be easily detected by instructors, so it's very important for cheaters to disguise their cheating and hew closely to an inexperienced writer, even a foreign-language speaker, in order to pass muster.
Also, you could probably get a ChatGPT instance to create and run a Python script to randomly replace letters in the words that ChatGPT itself creates. So technically, you can get it to misspell words. It seems like the person you replied to is forgetting the point of pre-training transformer models on instructions.
If a student doesn't engineer the prompt properly and the LLM uses its "native" voice, this can be easily detected by instructors, so it's very important for cheaters to disguise their cheating and hew closely to an inexperienced writer, even a foreign-language speaker, in order to pass muster.