Ask HN: How can we reliably verify content isn't AI generated?

AI generated media is rapidly becoming indistinguishable from human work. I feel the days of AI content feeling or looking 'off' are vanishing. With some AI generated videos like those from Veo 3 becoming so realistic, I'm already having to look twice, genuinely questioning if they're ai-generated or not.

How can we realistically maintain the distinction going forward? Is widespread, mandatory watermarking the only answer, or is it a technically flawed/doomed approach? And as AI content likely floods the internet, will we, and how could we, even try to filter or avoid it?