What is speakable schema?

Speakable schema is a schema.org property (speakable, with a SpeakableSpecification) that points, using CSS selectors or XPath, to the sections of a page best suited to being read aloud by a voice assistant, typically the headline and a short summary.

The speakable property can be added to an Article or WebPage in JSON-LD. Its value is a SpeakableSpecification with either a cssSelector array (for example the H1 and a paragraph with a given class) or an xpath array. The intent is that a text-to-speech assistant answering a query with your page reads only those passages, rather than the navigation, the byline and the cookie notice.

Google introduced support for it in 2018 for news publishers, aimed at Google Assistant and limited to English in the United States, and it has stayed in that beta state since. The honest summary is that adoption is thin, Google has never extended it to general business content, and no AI assistant has confirmed using it. It earns a place in an audit because it is cheap, it is harmless, and it forces a useful editorial habit: deciding which two or three sentences on a page are the answer.

For a business, that habit is the real value. If you can point a selector at a short passage that states what the page is about, that passage is also what a chat assistant is most likely to quote. An example is a case study page whose speakable selector targets the H1 and the summary paragraph directly beneath it, which is also the paragraph the answer-engine checks look for.

The common mistake is pointing the selector at the whole body, or at a class that a later redesign renames, so the markup references nothing. Keep the selectors simple and check them after every template change.

Related terms