Speakable Schema: Optimise Your Content for Google Assistant and Voice Search

Technically yes — you can add the markup to any page without errors. However, Google currently only processes Speakable schema for recognised news publishers (sites approved in Google News, using NewsArticle or Article schema types, in eligible languages: English, Portuguese, German, and French). For non-news sites, adding speakable markup today does not produce current Google voice results but may be processed by other voice assistants or future Google features. The more practical voice SEO strategy for non-news sites is FAQPage and HowTo schema.

Google recommends marking only 2–3 short sections per article as speakable. More than that, and you are asking Google to read a majority of your article aloud — which is impractical for voice delivery. The ideal speakable content is: (1) the headline/H1 (~10–15 words), and (2) a 2–3 sentence summary paragraph at the top of the article that directly answers the main question. Together these should take about 20–45 seconds to read aloud, which is the typical length of a useful voice search answer.

No — Speakable schema has no known effect on organic text search rankings. It is specifically for voice/audio output and AI content extraction. It does not cause pages to rank higher for desktop or mobile text search. There is a theoretical indirect benefit: if Speakable encourages AI Overviews to cite your content, that citation increases brand visibility. But this is speculative. There is no documented case of Speakable schema lifting a page's text search ranking position.

Both target specific HTML elements on your page for Google to read aloud. cssSelector uses CSS selector syntax (e.g. "#article-headline", ".summary-paragraph") — this is the recommended method as CSS selectors are simpler to write, widely understood, and reliably match elements in modern HTML. xpath uses XPath expression syntax (e.g. "/html/body/article/p[1]") — this was the earlier method, is more verbose, and is primarily useful for legacy CMSes that don't support id attributes or predictable class names on key elements. Use cssSelector unless you have a specific reason for xpath.

There is no direct "Speakable test" in Google's tools. Use these methods: (1) Google Rich Results Test — it detects speakable property presence and validates the schema structure, though it cannot simulate audio playback. (2) Google Search Console — check the URL Inspection tool to confirm the page's structured data includes speakable under the detected entities. (3) Ask Google Assistant directly about the topic your article covers — if your article is the featured source, the Assistant may read your speakable-marked sections. (4) Test using the Actions Console for Google Actions if you have a Smart Speaker Action.

Speakable was originally designed with AMP (Accelerated Mobile Pages) in mind for news publishers. On AMP pages, the cssSelector references must match the AMP HTML structure, which uses amp- prefixed components instead of standard HTML tags. The JSON-LD Speakable specification itself is identical — the difference is only in which element IDs and class names you reference. Both AMP and non-AMP pages are supported. Google has never confirmed that AMP pages have a Speakable advantage over standard pages.

Yes — if you update the article and the speakable sections' on-page content changes, you should update the cssSelector targets accordingly if the HTML structure changes. More importantly, if the factual summary (the speakable section) is no longer accurate after an update, update the on-page text in that section. Stale or inaccurate speakable content can negatively signal poor content freshness to AI systems. Always update dateModified in your NewsArticle schema when making significant updates.

Not directly — Speakable schema does not influence Google Discover eligibility. Discover eligibility is based primarily on: strong E-E-A-T signals, high-quality image (at least 1200px wide), topic freshness, and user engagement history. However, writing with voice-first clarity — short sentences, direct answers, no visual-only references — is the same writing style that tends to perform well in Discover. It is a writing discipline benefit rather than a schema signal benefit for Discover.