How should websites tell a helpful AI agent from a harmful one?
The “I’m not a robot” checkbox doesn’t really check anything, and bots now run real browsers. At the same time, more people are sending AI agents to shop and fill in forms for them.
Paul Graham said some startups will filter agents out and others will welcome them in. That fork is already here.
A few of us built a hackathon project around this and placed 3rd. I wrote up the problem in plain language, with stickman drawings: blog
What I’m curious about:
If you run a site, do you block agents, allow them, or not know yet?
Would you trust an agent to buy something for you? What would it need to prove first?
Is “are you human?” still the right question?
20 views
Replies
I run sites, and right now I'd lean toward letting them in. Visitors are already arriving on my sites from ChatGPT, so blocking every agent would mean blocking people who sent one on their behalf. The question I'd ask isn't whether it's human. It's whether whoever sent it is accountable for what it does.
Are you human isnt relevant anymore, by passing it is super simple and you are right we need to transition to fork model you are suggesting, but im not able to warp around head with a framework that can differentiate a good agent from a bad one, do we need to have passport for agents :0