Ok, fine, let's restrict it to AI agents only, without training. It's still an adversarial relationship with the content creator. When you take an AI agent an ask it "find me the best italian restaurant in city xyz" it scans all the restaurant review sites and gives you back a recommendation. The content creator bears all the burden of creating and hosting the content and reaps non of the reward as the AI agent has now inserted itself as a middleman.
The above is also a much clearer / more obvious case of copyright infringement than AI training.
> AI agent access cannot be banned without fundamentally breaking the User Agent model of the web.
This is a non-sequitur but yes you are right, everything in the future will be behind a login screen and search engines will die.
