The style wasn't too stiff. Their style was generic (no opinions, no nuance, no confidence) and they used a poor tool to check it. All LLMs use probabilities - at the most basic level, AI detectors predict what the next word in a sequence will be. If a text is very generic (and easy to predict), it can be flagged as AI. Good tools avoid this by combining other factors in their evaluation process.
Love it when AI detector owners reach out to notify us about significant updates so we can improve their listing ๐. Kudos @ai_detection! https://t.co/9qFWZxIKFn
@snoopysathena@zukolore If you don't write generic sentences, you shouldn't be. Remember - write with a confident tone, avoid words like "may, appers, should etc." and you will be fine. Good luck!
@Senshin108@TheQuillBot So-so. You can mitigate the risk by running the content through couple of good ones to get a overall understanding. Quillbot is OK but not the best. We agree that the final decision should be always human.
@epibrerehn@dirkjakehater Anything under 200 words is hard to analyze, so scanning messages for AI might not be the best thing to do. Good luck anyway!