Why a screenshot is not enough
Three things a picture hides, one limit that text does not solve either, and why Dinoer returns no image at all.
A screenshot shows how a page looks. It does not say what each element is, where a link leads, or what the page contains without displaying it. Dinoer produces no image, in any mode, and reads that information as text instead.
One: what an element is, and where it goes
On a screenshot, a link and a button styled the same way look the same, and so do a disabled control and an enabled one when the design uses a pale grey.
The page knows the difference: link "Learn more" /url: https://… gives the
role and the destination before anything is clicked. Looking at pixels does
not recover that. The accessibility tree →
Two: what is in the page but not displayed
A <dialog> that has not been opened is in the document, and its buttons
exist in the DOM. A collapsed menu holds its links; a hidden tab holds its
content.
Reading text instead of pixels does not solve this. The accessibility tree
only shows what is rendered: a closed <details> or <dialog> appears
without its content, just as a screenshot would show nothing of it. To see
what the document holds, read the DOM with evaluer
(document.querySelectorAll(…)), as the reconnaissance scenario does.
Three: whether the target is still the target
A screenshot is a moment. Between that moment and the action, a banner can close, a dialog can open, a list can gain rows. Anything identified by its position in that moment has moved: the usual way an automated click lands on the wrong element while reporting success.
Dinoer names every element by a CSS selector, never by a position on an image. A selector can still match another element after the DOM changes, but the position-based failure is gone.
So what is there to look at
Nothing.
If you need to see the page as a person does, open it in a browser. Dinoer returns text meant for deciding what to read or do next, not something to look at.
In short
- Roles and destinations have no visual form: that is the main reason to read structure rather than pixels.
- Present in the document is not the same as displayed, in text or in an image; read the DOM when it matters.
- Selectors name an element, not a position.
- There is no picture, for the model or for you.