What is Speakable Schema?
Speakable schema is a structured data type that marks specific sections of a webpage as the most suitable content for text-to-speech delivery — optimising it for voice assistants and AI reading tools. Implemented via CSS selectors or XPath in a Speakable JSON-LD block, it tells Google Assistant (and potentially other AI systems) exactly which sections of your page are the clearest, most concise answers appropriate for being read aloud. It's a direct signal for AEO and voice search optimisation.
- Speakable schema is currently limited to news articles in Google's official implementation — but signals voice-readiness to all AI systems.
- Mark only genuinely speakable content — concise, standalone sentences that make sense read aloud without surrounding context.
- Speakable sections should be 20–90 words — long enough to be informative, short enough for audio consumption.
- This is an early-stage schema type — implementing it now positions content for future AI audio interfaces.
- The content you mark as speakable should overlap with what you'd optimise for featured snippets — the formats align.
How Speakable Schema Works
Speakable schema uses the Speakable type in a JSON-LD block. You specify which page sections should be read aloud using CSS selectors (pointing to element IDs or class names) or XPath expressions.
When Google Assistant encounters a relevant query, it can retrieve the page and read aloud the marked speakable sections — providing an audio answer attributed to your site.
The markup structure: { "@context": "https://schema.org", "@type": "WebPage", "speakable": { "@type": "SpeakableSpecification", "cssSelector": [".article-headline", ".article-summary"] } }
Google's current implementation is focused on news publishers, but the underlying principle — identifying the most audio-appropriate content sections — applies broadly as voice and AI interfaces evolve.
What Content to Mark as Speakable
The best candidates for speakable markup: definition sentences (the first sentence of a definition, which directly explains the term), key takeaways (short, complete statements that stand alone), direct answers to common questions (the 40–60 word answer below an H2 question heading), and summary paragraphs.
Avoid marking: content that requires visual context (tables, charts), technical code examples, lists that don't make sense read linearly, and content with lots of parenthetical asides or complex sentence structures.
Think about the audio experience: 'Crawl budget is the number of pages Googlebot crawls on your site per day' reads well. 'As we discussed in the previous section, the following table shows...' does not.
Most guides are already outdated.
One email a week. The search stuff that actually matters — what shifted, what died, and what to do about it.
Subscribe free →The SpeakableSpecification type is formally proposed and added to the Schema.org vocabulary, establishing the technical foundation for identifying audio-appropriate content.
Google introduces Speakable schema in beta for Google Assistant, initially restricting eligibility to news publishers in the United States to test real-world audio retrieval at scale.
Google publishes official Speakable structured data guidelines, confirming CSS selector and XPath implementation methods and defining content quality criteria for marked sections.
The rise of AI-powered voice interfaces and large language model assistants increases the strategic value of Speakable markup as a direct signal for audio-optimised, machine-readable content selection.
| CSS Selector | XPath Expression |
|---|---|
| Simpler syntax — familiar to front-end developers and SEOs | More powerful — can target elements by content, position, or attribute value |
| Example: ".article-summary", "#page-intro" | Example: "/html/head/title", "//p[@class='summary']" |
| Best for modern, well-structured HTML with consistent class naming | Better suited to legacy CMS output or irregular DOM structures |
| Google's recommended and more commonly documented approach | Supported by Google but less commonly used in published implementations |
| Limited to element class and ID targeting | Can traverse parent-child relationships and conditional node selection |
How does your site score on AEO?
Paste your URL. Get a score and a fix list across all three disciplines. No form, no email.
Run Free Audit →Frequently Asked Questions
Not directly. Speakable schema is an AEO signal targeted at voice and audio interfaces, not a traditional ranking factor. However, the content it points to — concise, standalone, directly answerable — overlaps significantly with featured snippet and PAA content. Optimising for speakable content quality tends to improve all answer engine placements.
It's low-risk and potentially valuable as a future-proofing measure. AI reading tools, voice assistants, and text-to-speech interfaces are growing. Marking your best answerable content now costs little and signals content structure to AI systems beyond just Google Assistant. Validate with Google's Rich Results Test and implement on your most query-relevant pages.
- 1.Google — Speakable structured data documentation
- 2.Schema.org — Speakable type
- 3.Search Engine Land — Voice search schema
Read next
Conversational Search
Conversational search refers to search queries phrased as natural language questions or dialogue — the way peo…
Featured Snippets
A featured snippet is a selected search result that appears at the top of Google's results in a special box —…
How-To Schema
How-To schema is a structured data type (JSON-LD) that marks up step-by-step instructional content so Google c…
Schema Markup
Schema markup (structured data) is code you add to your web pages — typically in JSON-LD format — that explici…