Speakable schema: marking content for voice assistants
The speakable schema is a structured data type that tells voice assistants and smart speakers exactly which sections of a webpage are best suited for text-to-speech delivery. As answer engine optimization (AEO) reshapes how content is discovered and consumed, implementing speakable schema has become one of the most targeted schema markup strategies for answer engines available to publishers and SEO professionals alike.
How speakable schema works
The Speakable schema type, defined within the Schema.org vocabulary under SpeakableSpecification, allows webmasters to flag specific CSS selectors or XPath expressions that point to the most audio-friendly parts of a page. When a voice assistant processes a search query, it can use this markup to retrieve and read the most relevant passage aloud — rather than attempting to parse the entire page.
Core components of a speakable declaration
| Property | Type | Description |
|---|---|---|
| @type | SpeakableSpecification | Declares the speakable markup type |
| cssSelector | Text | Targets HTML elements by CSS class or ID |
| xpath | Text | Targets elements using XPath expressions |
Only one of the two selector properties — cssSelector or xpath — is required per SpeakableSpecification block. Using both in the same block is valid but redundant in most implementations.
Implementation syntax for speakable schema
JSON-LD example using cssSelector
JSON-LD is the recommended format for embedding speakable schema. The snippet below illustrates a minimal but complete implementation:
| JSON-LD field | Example value |
|---|---|
| @context | https://schema.org |
| @type | WebPage |
| speakable → @type | SpeakableSpecification |
| speakable → cssSelector | [“.speakable-intro”, “.key-answer”] |
| name | Page title |
| url | https://example.com/page |
Eligibility requirements
- ✅ Content must be original, editorial, and factual
- ✅ Pages should belong to recognized news or informational sites
- ✅ Marked sections must represent the core answer or summary — not navigational or promotional copy
- ✅ The page must be accessible to crawlers without login barriers
- ❌ Speakable markup on product pages, forms, or purely commercial content is not eligible
- ❌ Marking entire page bodies is discouraged; target only concise, self-contained passages
Writing principles for audio-friendly content
Speakable schema is only as effective as the copy it points to. Even perfectly valid markup will underperform if the flagged text is dense, jargon-heavy, or poorly structured for listening. Understanding the writing style principles that make speakable content effective is therefore just as important as the technical implementation itself.
✔ Do
- Use short, declarative sentences (under 25 words each)
- Answer the implied question in the first sentence
- Avoid abbreviations that are ambiguous when read aloud
- Use active voice consistently
- Keep paragraphs to 2–3 sentences maximum
✖ Avoid
- Tables, bullet lists, or visual-only structures inside speakable sections
- Parenthetical asides or em-dash-heavy sentences
- Relative references like “as shown above” or “click here”
- Marketing language or calls to action
- Nested or complex clause structures
Speakable schema within an AEO content strategy
Speakable schema does not operate in isolation. It works best as part of a broader structured data strategy — sitting alongside Article, FAQPage, and HowTo schema types — all of which feed into how answer engines surface and deliver content. Below is a practical mapping of where Speakable fits in the AEO ecosystem:
| Schema type | Primary use | Voice search role |
|---|---|---|
| Article | News and editorial content | Context and authorship signals |
| FAQPage | Question-answer pairs | Direct spoken answers to queries |
| HowTo | Step-by-step instructions | Sequential audio guidance |
| SpeakableSpecification | Highlighted text passages | Pinpointed audio-ready excerpts |
How Draftto generates speakable-compatible content
Producing content that meets both the structural and stylistic requirements of speakable schema is time-intensive when done manually. Draftto addresses this by generating answer paragraphs that are engineered for audio delivery from the ground up.
- 🎯 Draftto identifies the core query intent and opens every article section with a direct, concise answer sentence — the ideal candidate for a speakable CSS selector.
- 📐 Paragraph length and sentence complexity are automatically calibrated to voice-assistant reading standards, reducing editing time significantly.
- 🏷️ Draftto’s structured output assigns semantic CSS classes to key answer blocks, making it straightforward to reference those selectors directly in your JSON-LD speakable declaration.
- 🔗 Each generated article is aligned with the parent schema markup framework, ensuring speakable passages contribute coherently to the full AEO structured data layer.
For content teams managing high publication volumes, this means speakable schema implementation stops being a post-production task and becomes a natural output of the content generation workflow itself.
Validating and monitoring speakable markup
Testing tools and signals to track
- Use Google’s Rich Results Test to confirm that speakable markup is parsed correctly
- Check Google Search Console for structured data errors related to SpeakableSpecification
- Monitor voice search impressions through third-party voice analytics platforms
- Audit flagged sections periodically to ensure the copy remains concise and factually accurate
- Verify that CSS selectors in the JSON-LD continue to match live page elements after template or CMS updates
Speakable schema represents one of the most direct bridges between written content and spoken search results. When paired with clean, audio-optimized copy and integrated into a complete speakable schema and AEO structured data strategy, it positions content to be retrieved and read aloud by voice assistants with precision — making it an essential markup type for any publisher serious about answer engine visibility.