Speakable schema: marking content for voice answer surfaces
SpeakableSpecification is a JSON-LD type that marks specific CSS selectors as suitable for text-to-speech. Voice assistants and AI answer surfaces use it to pick a sentence to read aloud — useful for news, definitions, and short factual answers.
SpeakableSpecification is a JSON-LD type that marks specific CSS selectors as suitable for text-to-speech. Voice assistants and AI answer surfaces use it to pick a sentence to read aloud — useful for news, definitions, and short factual answers. Inside Article or NewsArticle schema, you add a speakable property with cssSelector entries pointing at the part of the page that contains the answer. Currently a beta feature in Google but already used by some voice products.
Last updated·
What it is
Inside Article or NewsArticle schema, you add a speakable property with cssSelector entries pointing at the part of the page that contains the answer. Currently a beta feature in Google but already used by some voice products.
Why it matters
For factual / news / definition queries, voice surfaces (Google Assistant, Alexa news briefings) lean on speakable when present. Marking the right sentence wins voice answer position; not marking anything yields the default first paragraph.
How to fix it
- Add speakable to Article JSON-LD. "speakable": {"@type": "SpeakableSpecification", "cssSelector": [".lede", "h1"]}. Point at the part of the page that answers the implied query.
- Choose the lede as the speakable selector. Your one-sentence direct answer is usually the right thing to read aloud. Mark its CSS class.
- Validate with Rich Results Test. Google's tester surfaces speakable in supported article types.
Authoritative sources
- Google Search Central documentation — Google
- Schema.org vocabulary — schema.org
- SEO Starter Guide — Google Search Central
- MDN — HTML meta and link elements — Mozilla MDN