Extraction

Jint.Browser provides text-first alternatives to screenshots:

string markdown = await page.MarkdownAsync(mainContentOnly: true, maxLength: 4_000);
string text = await page.TextAsync();
string ax = await page.AccessibilitySnapshotAsync(includeReferences: true);

• MarkdownAsync returns CommonMark.
• TextAsync returns document text with structural line breaks and whitespace processing, but no visual wrapping.
• AccessibilitySnapshotAsync returns an indented outline of roles, accessible names, states, and text.

mainContentOnly selects the first <main>, [role=main], or <article> when one exists. maxLength of 0 means unlimited. A truncated result ends at a word boundary and includes [truncated].

Extraction reads the DOM and runs no page JavaScript. It is computed on demand and does not maintain a rendered representation.

Accessibility references make a snapshot actionable:

- button "Save" [ref=42]

Input methods accept ref=42 in place of a selector. References remain valid only for the current document.

The accessibility tree cannot determine whether an element is off-screen, clipped, or covered because there is no visual layout. Visibility is based on document and CSS information the package can compute.