Speakable Schema
Speakable schema is structured data, using the schema.org speakable property, that points to the parts of a web page best suited to being read aloud by a voice assistant. Google introduced it as a beta feature for news publishers, so its practical use is narrow, but it shows how content can be prepared for audio answers.
How speakable schema works
Speakable is a property added to Article or WebPage schema markup. Instead of repeating the text, it identifies which sections of the page are speakable, using either:
- cssSelector: CSS selectors, such as a class on a summary paragraph.
- xpath: XPath expressions that locate elements in the HTML.
A voice assistant can then use text-to-speech to read those sections when someone asks a question that the article answers. It is usually added as JSON-LD.
Who can use it
Google has listed speakable as a beta feature with limited availability. Its documentation has described it as aimed at news publishers, for English-language users in the United States on Google Assistant devices. Because of these limits, most businesses will see no visible effect from adding it. If you publish news, check Google’s current structured data documentation before relying on it, as beta features can change or be withdrawn.
Why it still matters as an idea
Even if you never add the markup, the principle behind it is useful for voice search and AI answers. Voice assistants and AI tools prefer short, self-contained passages that make sense when heard or quoted without the rest of the page. Writing a clear summary near the top of an article helps whether or not the summary is marked up.
How to write speakable content
- Keep each speakable section short. Google’s guidance suggested passages of roughly 20 to 30 seconds when read aloud, which is about two or three sentences.
- Lead with the key fact or answer, rather than background.
- Avoid content that does not work as audio, such as captions, tables, datelines and source credits.
- Write in plain sentences that do not rely on visual layout or links to make sense.
- Make sure the marked sections are visible on the page and match what the markup points to.
Speakable versus other answer formats
Speakable content overlaps with other answer formats. The same concise paragraph that suits audio often works as a featured snippet or a quotable passage for AI systems.
Thinking in terms of short, extractable answers supports answer engine optimisation more broadly, and connects to how conversational search tools respond to spoken questions.
Common mistakes
- Adding speakable markup to non-news pages and expecting a change in results.
- Pointing selectors at elements that do not exist or that change between templates.
- Marking whole articles as speakable instead of short summaries.
- Choosing sections full of numbers, abbreviations or references that sound confusing when read out.
Further reading: Google’s speakable structured data documentation.
Key Takeaways
- Speakable schema flags sections of a page suited to text-to-speech playback.
- It uses CSS selectors or XPath to point to the relevant text.
- Google has offered it as a limited beta aimed at news publishers.
- Most sites will see no visible effect from adding it.
- The underlying idea of short, self-contained answers helps voice and AI search.
Related Terms
Curious how this applies to your website? Request a free SEO check-up, or keep learning in the SEO Encyclopedia.