Definition
Speakable is a Schema.org property that marks the sections of an article or web page best suited to being read aloud by text-to-speech. Publishers point to those sections with a SpeakableSpecification that uses either CSS selectors or XPath expressions, never both. Google introduced support in 2018 for news content read by Google Assistant, and its speakable documentation still labels the feature BETA in October 2026.
Google's use case is narrow. According to that documentation, Google Assistant uses speakable data to answer topical news queries on smart speakers, returning up to three articles and reading the marked sections aloud, then sending the full article link to the user's phone. It works only for users in the US with Google Home devices set to English, and only for English-language publishers. Google recommends about 20 to 30 seconds of audio per section, roughly two or three sentences, and advises against marking datelines, photo captions or source attributions.
That consumer is changing. Google is moving its speakers and displays from Google Assistant to Gemini for Home, which is rolling out through early access and ships on the Google Home Speaker (2026). Google's help pages say compatible devices switch permanently once a household opts in. Google has not said whether Gemini for Home reads speakable markup, and the speakable documentation still refers only to Google Assistant.
Speakable is not a signal for Google's AI features. Google's May 2026 guide to generative AI in Search says no special Schema.org markup is needed for AI Overviews or AI Mode. Voice modes in ChatGPT, Gemini Live and similar assistants read their own generated answers aloud rather than playing marked sections of your page, and none of them documents speakable as an input.
Where speakable still fits:
- English-language news publishers in the US that want eligibility for Google Assistant news readouts while the feature exists.
- Teams that use the markup internally to tag summary sections for their own audio or text-to-speech products.
The writing principle behind speakable is more durable than the markup. A two or three sentence summary that makes sense out of context, with no 'as shown above' or visual references, is the kind of passage that voice assistants and AI answers can lift cleanly. That overlaps with answer-ready content and the advice under voice search optimization. For most sites outside news publishing, writing those passages matters and the markup does not. See schema markup for which structured data types Google still turns into visible features and voice search for how spoken queries differ from typed ones.
Examples of Speakable
- A US news site marks each article's headline and two-sentence summary with a cssSelector-based SpeakableSpecification, so Google Assistant can read the summary when someone asks for the latest news on a topic.
- A B2B software company removes speakable markup that an SEO plugin added to every blog post, after confirming it has no effect outside Google Assistant news readouts.
- A public broadcaster uses the same 'speakable' CSS class to choose which paragraphs its own text-to-speech player reads, whether or not Google consumes the markup.
- A local health service rewrites its clinic pages so each starts with a short, self-contained answer about opening hours and walk-in rules, which reads well in voice and AI answers without any markup.
