tbh i've heard mixed things. a lot of those ai detectors, including safeassign's underlying tech, aren't perfect. sometimes it flags genuinely human writing, sometimes it misses obvious ai. i'd say if you just paste raw chatgpt, it's pretty risky. but if you rephrase and edit it yourself, it's much harder for it to detect. it's less about 'chatgpt' specifically and more about patterns in the text.
