How AI generated text is about to be exposed

Systems to detect when AI has been used to write an assignment, for example, are already being used at schools and universities. But there's a change underway at some of the biggest artificial intelligence firms that could make detectors much harder to fool. Watermarks are coming to text generated by AI chatbots. But you can't see it and you probably won't be able to tell if it's been applied. How does it work?