Detecting AI content will always be hard. To do it at scale, you’d need to use AI, there’s basically no alternative, especially for search engines. Thr issue is AI models are constantly training to avoid detection, so we will always be in a state where AI cannot reliably detect AI output.
Ai images and movies can be invisible watermaked in image, sound or just metadata. Gemini is already doing that. It would not be hard to look for those signals during upload. YouTube is already doing something similar when it checks your videos for copyrighted music, content from other creators or advertiser unfriendly no-no words.
The users arent the ones that need managed. Its the creators. And they will absolutely know how to remove marking metadata. Thats why its valled an Arms “Race”. Its a constant need that is ever adapting.
Detecting AI content will always be hard. To do it at scale, you’d need to use AI, there’s basically no alternative, especially for search engines. Thr issue is AI models are constantly training to avoid detection, so we will always be in a state where AI cannot reliably detect AI output.
Ai images and movies can be invisible watermaked in image, sound or just metadata. Gemini is already doing that. It would not be hard to look for those signals during upload. YouTube is already doing something similar when it checks your videos for copyrighted music, content from other creators or advertiser unfriendly no-no words.
Stripping metadata is not hard. If that gets you into the human section of the web, it’s absolutely worth doing.
99% percent of users won’t know what it is, or how to remove it.
The users arent the ones that need managed. Its the creators. And they will absolutely know how to remove marking metadata. Thats why its valled an Arms “Race”. Its a constant need that is ever adapting.