Speakable Schema: What It Does in 2026

Speakable schema is structured data that marks the sections of an article or web page best suited to being read aloud by text-to-speech, used by Google Assistant for news in a beta limited to U.S. English.

What is speakable schema?

Speakable schema is structured data that points to the parts of a page that work best when read aloud. A news publisher adds the speakable property to its Article or WebPage markup and lists the headline and summary, and a voice assistant can then read those sections to a listener using text-to-speech (TTS) ^[1].

{
  "@context": "https://schema.org/",
  "@type": "WebPage",
  "name": "Quarterly jobs report shows hiring slowed",
  "speakable": {
    "@type": "SpeakableSpecification",
    "cssSelector": [".headline", ".summary"]
  },
  "url": "https://www.example.com/jobs-report"
}

Speakable schema does not change how a page looks or ranks in ordinary search results. Speakable schema only tells a voice system which text to speak. Google's search gallery describes it as a way to "identify news content to read aloud on Google Assistant-enabled devices" ^[2].

How does speakable schema work?

Speakable schema works by pairing a page with selectors that pick out the text to read. When someone asks Google Assistant on a smart speaker for news on a topic, Google Assistant can return up to three articles from around the web and read aloud the sections each publisher marked as speakable ^[1].

Google Assistant also credits the source when it reads a speakable section, and sends the full article URL to the listener's phone through the Google Assistant app ^[1]. That handoff is the only traffic route speakable schema offers: a listener who wants more opens the link later.

Who can use speakable schema?

Speakable schema can be added by any site, but Google only uses it for a narrow audience. Google's documentation says the speakable property works for users in the U.S. with Google Home devices set to English, and for publishers that publish content in English ^[1].

That limits speakable schema in three ways:

Content: news. Google describes the feature as audio playback for "topical news queries" ^[1].
Market: U.S. English only ^[1].
Device: Google Home and other Google Assistant smart speakers, not ordinary Google Search ^[1].

A business in Australia, the UK or Canada, a store selling products, or a site publishing evergreen guides gets no documented benefit from speakable schema in Google.

How do you add speakable schema to a page?

To add speakable schema, add a speakable property of type SpeakableSpecification to the page's Article or WebPage JSON-LD, then point it at the speakable text with either cssSelector or xPath ^[1]. Google requires one of the two.

With CSS selectors, mark the speakable elements with classes and list them:

"speakable": {
  "@type": "SpeakableSpecification",
  "cssSelector": [".headline", ".summary"]
}

With XPath, point at elements by their position in the HTML, such as the title and meta description:

"speakable": {
  "@type": "SpeakableSpecification",
  "xPath": [
    "/html/head/title",
    "/html/head/meta[@name='description']/@content"
  ]
}

Both examples follow Google's documentation ^[1]. CSS selectors are usually safer, because a template change can silently break an XPath that depends on exact page structure.

What content should speakable schema mark?

Speakable schema should mark short, self-contained text that makes sense heard aloud with no visual context. Google recommends sections of about 20 to 30 seconds of audio, roughly two or three sentences, and short sentences at the top of the story ^[1].

Google also says not to mark text that sounds wrong when spoken ^[1]:

Datelines such as "WASHINGTON (Oct. 2)."
Photo captions, which describe images the listener cannot see.
Source attributions and other labels meant for the eye.

The aim is to mark the key points, not the whole article ^[1]. In practice that usually means the headline plus a two-sentence summary written to be heard: full names instead of pronouns, no "see chart below," and numbers written so they read naturally.

What is the difference between speakable schema and FAQ or HowTo markup?

Speakable schema marks text to be read aloud by a voice assistant, while FAQ and HowTo markup describe content so Google can show it as a visual rich result on the search results page. Speakable schema never produced a visible search feature; its only documented output is audio on Google Assistant devices ^[1]^[2].

The two also differ in their fate. Google has spent 2024 and 2025 removing structured data features that no longer appear in search results, retiring course info, estimated salary, learning video, special announcement and vehicle listing markup in September 2025, and phasing out practice problem and dataset markup in November 2025 as part of simplifying the results page ^[3]. Speakable is still listed in Google's structured data gallery and still documented as a beta feature as of October 2, 2026 ^[1]^[2].

Does speakable schema help voice search or AI answers?

There is no published evidence that speakable schema helps voice search beyond Google Assistant news playback, or that it helps AI answers at all. Google documents one use: reading marked news sections aloud on Google Home devices in U.S. English ^[1]. Google has not said that AI Overviews, AI Mode or Gemini read the speakable property.

Many guides go further and present speakable schema as a general way to "optimize for voice search" or to get quoted by AI assistants. Those are claims, not documented behavior. Writing short, self-contained summary passages does help pages get quoted, but that comes from the writing, not the markup. The speakable property is a pointer to that text, not a signal that makes it more quotable.

Has Google changed or retired speakable schema?

Google has not retired speakable schema: it remains a documented beta feature ^[1]. The wider Google Assistant developer platform has changed around it, though.

2022 to 2023: Conversational Actions sunset

Google sunset Conversational Actions, the custom voice apps developers built for Google Assistant, with a turndown on June 13, 2023 after a year-long sunset period that began June 13, 2022 ^[5]^[4]. Google listed "Actions from web content" among the Assistant development paths that continued unaffected ^[5]. Speakable schema is web-content markup, so the sunset did not remove it.

2025: Google prunes structured data features

Through 2025 Google removed documentation for several structured data types that no longer showed in results ^[3]. Speakable was not among them, and it remains in the gallery ^[2].

When should you use speakable schema?

Use speakable schema if you publish English-language news for a U.S. audience and want a chance at being read aloud by Google Assistant on smart speakers ^[1]. For that publisher it is cheap to add: most CMS templates already have a headline and summary that can be selected with two classes.

Skip speakable schema, or treat it as optional, if:

You are not a news publisher. Google ties the feature to topical news queries ^[1].
Your audience is outside the U.S. or not English-speaking ^[1].
You expect it to improve rankings or AI citations. No source shows that it does.

Whatever you decide about the markup, the writing habit it encourages is worth keeping: open each article with a two or three sentence summary that a listener, a reader skimming on a phone, or an AI system quoting a single passage could understand on its own.

Speakable Schema