Now I know how the Salem witches felt. How can I prove that it's not AI?

You can't. Nobody can. False positives are the inherent danger of these sorts of policies — especially when the LLMs were trained on the exact writing styles that have dominated online conversations and publications for decades.