How to use Speakable structured data for voice search

What is speakable structured data?

Speakable structured data is a schema.org property that identifies specific sections within an article or webpage particularly suitable for audio playback using text-to-speech (TTS) conversion. Its primary purpose is to highlight content that is especially appropriate for this conversion. While other parts of a page might also be spoken, the `speakable` property indicates the sections most likely to be generally useful for speech.

This markup allows search engines and other applications to identify content that can be read aloud on voice-enabled devices. For example, Google documents that its Assistant uses `speakable` structured data to answer topical news queries on smart speaker devices. This helps make your content accessible to users interacting through voice interfaces.

Why should you use speakable structured data?

Using `speakable` structured data helps your content reach users through voice interfaces. Google documents that this markup allows search engines and other applications to identify content to read aloud on Google Assistant-enabled devices. Specifically, the Google Assistant uses this data to answer topical news queries on smart speaker devices.

The `speakable` property helps highlight the parts of a page that are most likely to be generally useful for speech, improving the overall audio user experience. Google recommends around 20-30 seconds of content per section, or roughly two to three sentences, for optimal audio experiences.

What content should you mark as speakable?

When applying `speakable` structured data, Google recommends focusing on key points of your content rather than highlighting an entire article.

It is important to avoid adding `speakable` structured data to content that might sound confusing in voice-only or voice-forward situations. Examples of content to exclude are datelines, photo captions, or source attributions. For the best audio user experiences, Google recommends that each section of `speakable` structured data contain approximately 20-30 seconds of content, which typically translates to about two to three sentences.

How do you implement speakable structured data?

The `speakable` property can be repeated multiple times within a page to mark various sections. Schema.org specifies that there are three kinds of possible "content-locator" values you can use to identify the content: ID-value URL references, CSS selectors, and XPaths. Google also documents that CSS selectors and XPaths are possible content-locator values.

For simple cases, you can use an ID-value URL reference, which points to the `id` attribute of an element on the page being annotated. For more complex or sophisticated markup of speakable sections, Schema.org recommends using either CSS selectors or XPath expressions. CSS selectors address content, for example, via a class attribute, while XPaths address content assuming an XML view of the document.

Schema.org defines a supporting type called `SpeakableSpecification`, which is a possible value for the `speakable` property. Instances of `SpeakableSpecification` are primarily used to indicate sections of a document that are particularly speakable, typically via its `xpath` or `cssSelector` properties. The `cssSelector` property takes a CSS selector, and the `xpath` property takes an XPath.

How do you check your speakable structured data?

After implementing `speakable` structured data, it is important to validate its correct application. Google recommends using its "Troubleshoot missing rich results / drop in total rich results" guide. This resource provides a step-by-step approach to identify, fix, and validate any issues with your structured data within Search Console. Regular checks help ensure your content is correctly interpreted for voice search.

Common questions