Speakable

The Speakable property is a Schema.org type that pinpoints specific parts of an article or webpage suitable for text-to-speech (TTS) conversion. It uses XPath or CSS selectors to target HTML elements. When a voice assistant receives a query, it can parse the structured data to extract only the speakable content, bypassing navigation, ads, and tangential text. This results in a direct, audible answer rather than a robotic reading of the entire page.

Implementation requires embedding a SpeakableSpecification within a WebPage or Article schema. The xpath or cssSelector property points to the DOM node containing the speakable text.

• JSON-LD Example: "speakable": { "@type": "SpeakableSpecification", "xpath": ["/html/head/title", "/html/body/div[2]/p[1]"] }
• Content Rules: The targeted text must be straightforward, concise, and grammatically correct when read aloud. Avoid complex tables or nested lists.
• Validation: Use Google's Rich Results Test to confirm the markup is correctly parsed.

Content marked as speakable must adhere to strict editorial guidelines to ensure a natural auditory experience:

• Brevity: Limit to 20-30 seconds of spoken audio (approx. 2-3 sentences).
• Clarity: Use simple sentence structures. Avoid jargon that is hard to pronounce.
• Self-Contained: The snippet must make sense without visual context or surrounding paragraphs.
• No Visual Cues: Avoid phrases like "click here" or "as seen in the chart below."

While primarily an SEO tool for voice assistants, speakable complements Web Accessibility Initiative (WAI) standards. Screen readers can theoretically leverage this markup to skip repetitive UI elements and jump to the core content. However, speakable is not a replacement for ARIA labels or proper semantic HTML. It serves as a parallel, AI-focused signal that reinforces the primary narrative of the page for auditory consumption.

The speakable property has specific technical constraints:

• Eligibility: Currently, Google primarily supports it for news articles and blog posts.
• No Markup in Target: The selected HTML element should contain only text. Inline markup like <em> or <a> tags can cause TTS engines to stumble.
• Single Focus: Do not mark up the entire body. Select only the most newsworthy summary sentence.
• Testing: Always listen to the output via a TTS simulator to catch awkward pronunciations before deployment.

The parent Schema.org type under which Speakable is declared. A WebPage entity represents a single document on the web and serves as the container for properties like speakable, mainEntity, and breadcrumb.

• Speakable is a property of WebPage (and its subtypes like Article, NewsArticle)
• Declaring a WebPage with @id enables cross-referencing within a knowledge graph
• Voice assistants use the WebPage context to determine which sections to prioritize for TTS output

A Schema.org property that identifies the primary topic of a webpage, helping search engines and voice assistants disambiguate the central subject from peripheral content.

• Works in tandem with Speakable: MainEntity defines what the page is about, Speakable defines which text to read aloud
• Critical for voice search accuracy—assistants use MainEntity to confirm relevance before reading speakable sections
• Example: On a product page, MainEntity points to the Product type while Speakable highlights the description paragraph

A Schema.org type for marking up glossary entries and definitions with formal term-description pairs. When combined with Speakable, it enables voice assistants to deliver concise dictionary-style answers.

• Use DefinedTerm within a DefinedTermSet to structure entire glossaries
• Speakable can point to the description property of a DefinedTerm for TTS-friendly definition snippets
• Enhances eligibility for dictionary-rich results and voice answer boxes

A Schema.org type that structures question-and-answer pairs on a single page. Each Question and Answer block becomes a discrete unit that voice assistants can parse and read aloud sequentially.

• Speakable can target individual Answer nodes within an FAQPage for precise TTS selection
• Google often surfaces FAQPage markup as rich results and voice answers
• Best practice: Keep each Answer concise (under 2-3 sentences) for optimal voice delivery

A Schema.org type used to fact-check a specific statement, indicating the claim, the review verdict, and the fact-checking source. When marked as speakable, voice assistants can deliver verified information with explicit credibility signals.

• Speakable applied to ClaimReview enables voice assistants to read fact-check verdicts directly
• Builds algorithmic trust—AI models weigh speakable, fact-checked content as high-confidence sources
• Critical for news publishers combating misinformation in voice search results