Speakable Schema for News Publishers: Enabling Voice Delivery of News Content
Speakable schema (the SpeakableSpecification structured data type on Schema.org) enables news publishers to explicitly mark which sections of their articles should be delivered via voice interfaces - primarily Google Assistant voice responses and Google News audio summaries. Introduced as a beta feature for news publishers, Speakable is the only structured data type that creates a direct bridge between written news content and AI voice delivery: a user asking Google Assistant 'What's happening with [topic]?' triggers Speakable-marked excerpts from top news sources to be read aloud.
Unlike FAQPage and HowTo schema which are available to all websites, Speakable is currently restricted to Google News publishers. For eligible publishers, implementing Speakable schema transforms their articles from text-only citations into voice-deliverable audio content - a significant expansion of distribution reach as smart speaker usage and voice AI query volume continue growing. The implementation is technically simple (CSS selectors targeting existing HTML elements) but content requirements are strict: Speakable sections must be concise, answer-first, factual, and entirely free of promotional language.
For context on broader voice schema strategy, see Speakable Implementation, NewsArticle Schema, and Voice Featured Snippets.
Speakable for News - End-to-End Voice Delivery Flow
From article publication through Google voice delivery - five stages that explain what happens after Speakable schema is added:
News Article Published: Journalist publishes a news story. Key summary paragraphs exist that answer the 'what happened' question for voice.
cssSelector vs xPath - Complete Implementation Examples
Both cssSelector and xPath methods are supported. For most CMS implementations, cssSelector is more maintainable - select your method and see the full annotated JSON-LD example:
cssSelector (Recommended for most CMS)
{
"@context": "https://schema.org",
"@type": "NewsArticle",
"name": "AI Overviews Expand to 80 Countries",
"datePublished": "2026-03-13",
"author": { "@type": "Person", "name": "Jane Smith" },
"speakable": {
"@type": "SpeakableSpecification",
"cssSelector": [
".article-summary", // Main summary paragraph
".article__lede", // Lede/standfirst paragraph
"h1", // Article headline
".key-points li" // Bulleted key points list
]
}
}Use CSS class selectors for CMS content - cleaner and more template-stable than xPath
Multiple selectors array: each selected element's text is included in the speakable excerpt
Avoid positional CSS selectors like 'p:nth-child(1)' - template changes can break them
Test that selected elements exist and contain voice-appropriate text using Google's Rich Results Test
Speakable Content Requirements - Specifications and Restrictions
Exact technical specifications, content format requirements, and restrictions for Speakable eligibility:
Content type restriction
Google's Speakable schema is currently only supported for news publishers enrolled in Google News. Non-news content sites cannot use Speakable for voice assistant delivery.
Optimal excerpt length
Speakable sections should be 20–120 words per selector target. Voice assistants read approximately 1 minute of audio maximum. Multiple short sections are better than one long paragraph.
Answer-first format required
Speakable sections must deliver the news fact immediately. 'Today, Google announced...' is ideal. 'In order to understand the significance of this announcement...' will not be selected for voice delivery.
No promotional content
Speakable markup on sections containing advertising copy, subscription prompts, or promotional language will fail validation and may result in the entire page being excluded from voice delivery.
No navigation or boilerplate
Do not target headers, footers, menus, bylines, or publication date stamps as speakable content. Target only the article body's factual summary content.
Multiple sections per article
Two to four targeted speakable sections per article is optimal. Single selectors underutilize the capability; more than four sections may indicate the excerpt is too broad to be voice-appropriate.
Validation requirement
Test implementation at search.google.com/test/rich-results before publishing. Speakable schema that fails validation produces no voice delivery benefit and may emit Search Console warnings.