A website is now also an interface for software

For most of the web's life a page had two audiences: a person, and a crawler that wanted to describe the page to a person. There is a third one now, and it does not want to describe your page. It wants to do something on it.

What actually changes

Less than the noise suggests, and in one specific direction: the cost of a change drops.

Small businesses do not usually stop maintaining a website because they cannot design one. They stop because a change costs a phone call, a quote and a week. If describing the change is enough to make it, the site stays current, and a current site is worth more than a beautiful one.

What should not change

Two things are worth protecting.

The first is that a person still has to be able to open the thing and edit it. Anything an agent builds here is ordinary blocks in the ordinary editor. If the tooling went away tomorrow, the site would not.

The second is that permission stays where it was. An agent acts as you, inside your account, with the access you already had. A tool that skipped a check to be convenient would be a bug, not a feature.

An account, a starting point, and a form that sends you a real email. Nothing is public until you say so.