OpenAI will watermark ChatGPT outputs by default—but only in the EU (arstechnica.com)

Like other solutions, it is not especially reliable, and it's easy to circumvent.