What I mean is that if I entered a sentence into ChatGPT/Gemini/Grok and tell it to fix the flagged word, it will be able to get it right almost all of the time (assuming it's not a weird proper noun or inside joke slang).