HTML Living Standard

Unlike the W3C's versioned HTML5 approach, the WHATWG maintains the HTML Living Standard as a continuously updated document. There is no 'HTML6'—new features, security patches, and parsing refinements are integrated directly into the standard as they achieve consensus. This ensures the specification always reflects browser implementation reality rather than theoretical milestones.

The standard defines an exhaustive, deterministic parsing algorithm that dictates exactly how every byte stream must be converted into a DOM tree. This eliminates ambiguity:

• Error handling is fully specified—browsers must recover from malformed markup identically
• AI crawlers and headless parsers can replicate browser DOM construction precisely
• Ensures semantic extraction is consistent across all conforming user agents

The standard catalogs every valid HTML element with its:

• Content categories (Flow, Phrasing, Sectioning, Metadata, Embedded)
• Permitted parent-child relationships
• Required ARIA implicit roles
• Allowed ARIA roles, states, and properties

This vocabulary forms the foundation of programmatic determinism, enabling AI agents to infer meaning from element choice alone.

The standard establishes the Document Object Model (DOM) as the definitive representation of a document. All APIs, CSS rendering, and accessibility tree construction operate on the DOM, not the raw HTML source. For AI systems, this means:

• JavaScript-modified content is the canonical state
• The accessibility tree is derived directly from the DOM
• Shadow DOM boundaries define encapsulation scopes for custom elements

The standard mandates indefinite backward compatibility. Deprecated elements like <center> or <font> are documented as obsolete but must still be parsed and rendered consistently. This ensures:

• Legacy web content remains accessible to modern AI crawlers
• Parsing rules for archaic markup are explicitly preserved
• New features are layered on without breaking existing document interpretation

The HTML Living Standard explicitly defers to and integrates with:

• WAI-ARIA for accessibility semantics beyond native HTML
• Microdata and JSON-LD for embedded structured data
• Fetch, URL, and Encoding standards for network and text processing
• Web IDL for defining JavaScript API bindings

This modular architecture allows AI systems to rely on a coherent stack of interoperable specifications.

A logical, nested structure of HTML heading elements (<h1> through <h6>) that defines the document outline. This communicates the relative importance and parent-child relationships of content sections to search engine parsers and accessibility bots.

• <h1>: The single, primary topic of the page
• <h2>: Major sections subordinate to the <h1>
• <h3>-<h6>: Increasingly granular subsections
• Critical Rule: Never skip levels (e.g., <h2> directly to <h4>)
• AI models use heading hierarchy to weight content importance and generate accurate summaries

A set of predefined roles added to HTML elements to programmatically identify distinct page regions. These allow assistive technologies and AI agents to efficiently navigate and understand document layout without visual parsing.

• Common Landmarks: banner, navigation, main, complementary, contentinfo, search, form, region
• Can be applied via HTML5 semantic elements (implicit) or explicit role attributes
• Example: <header role="banner"> or simply <header> when not nested
• AI crawlers use landmarks to isolate high-value content zones from boilerplate like headers and footers

A parallel structure generated by the browser from the DOM that exposes semantic information, properties, and relationships of UI elements exclusively to assistive technologies and programmatic agents.

• Contains accessible names, roles, states, and values for every meaningful node
• Stripped of purely visual CSS and non-semantic markup
• AI crawlers and screen readers query this tree, not the visual render
• A malformed accessibility tree (caused by Divitis or missing ARIA) directly degrades AI content extraction accuracy

The hierarchical, object-oriented representation of an HTML document's parsed markup. This forms the node tree that scripts and AI crawlers traverse to extract, manipulate, and understand content relationships.

• Node Types: Element nodes, text nodes, attribute nodes, comment nodes
• A flat, deeply nested, or invalid DOM complicates AI parsing
• Shadow DOM encapsulates component internals but requires declarative techniques to expose semantics
• The DOM is the single source of truth that the browser, JavaScript, and AI agents all interact with

An HTML5 specification for nesting machine-readable metadata within existing content using itemscope, itemtype, and itemprop attributes. This provides a direct annotation mechanism for AI parsers to extract structured entities without relying on separate JSON-LD blocks.

• itemscope: Declares a new structured data item
• itemtype: Specifies the vocabulary (e.g., https://schema.org/Person)
• itemprop: Labels a property of the item (e.g., name, jobTitle)
• Unlike JSON-LD, Microdata is embedded directly in the visible markup, reducing the risk of content-model mismatch
• Still actively parsed by major search engines alongside JSON-LD