Speakable Schema

Why It Matters

Voice search and AI audio responses require content that sounds natural when spoken aloud-not every web page passage translates well to audio. Speakable schema enables publishers to explicitly designate their best audio-appropriate content, increasing the probability that AI voice assistants select their content for audio playback. As smart speakers and AI voice interfaces grow, speakable markup becomes a meaningful voice search optimization lever.

How It Works

Speakable schema is implemented as JSON-LD using the Speakable type within NewsArticle or WebPage schemas, specifying which CSS selectors or XPath expressions identify the speakable content sections. Google's text-to-speech systems and voice assistant platforms can use these designations to identify optimal content sections for audio synthesis and delivery.

Use Cases

• News publishers marking article lead paragraphs as speakable for Google Nest and voice assistant news briefings
• FAQ pages marking direct answer sections as speakable for voice search response delivery
• Recipe sites marking ingredient lists and step summaries as speakable for kitchen voice assistant use
• Financial news publishers marking market summary sections as speakable for voice briefing services
• Health information sites marking symptom and treatment summaries as speakable for voice health queries

Best Practices

• Mark only concise, factually dense sections as speakable-long narrative passages don't translate well to audio
• Ensure speakable sections contain complete, standalone answers understandable without visual context
• Target 20–30 second audio equivalents (approximately 50–75 words) for each speakable section
• Validate speakable implementation using Google's Rich Results Test
• Prioritize speakable implementation for content targeting common voice query patterns
• Write speakable sections in natural spoken language-avoid visual formatting elements like bullet points