Glossary
Speakable Schema
Speakable schema is a Schema.org specification that identifies the sections of a web page most suitable for text-to-speech playback and AI voice assistant responses. It uses CSS selectors or XPath expressions to mark content that is concise, informative, and well-suited for audio delivery, helping voice assistants and AI models select the best content to read aloud.
What Speakable Schema Is
Speakable schema is part of the Schema.org vocabulary designed for the age of voice assistants and AI-generated answers. It tells machines: if you need to read part of this page aloud or extract a quotable summary, these are the best sections to use.
The specification uses a SpeakableSpecification type with cssSelector or xpath properties that point to specific DOM elements. This is more targeted than marking an entire page as speakable; it guides AI systems to the exact paragraphs or sections that work best in an audio or citation context.
Why It Matters
Voice search and AI assistants need concise, well-structured content to deliver good responses. Without speakable markup, these systems must guess which parts of a page to read or cite. They might select a navigation element, a disclaimer, or a verbose paragraph that does not represent your core message.
Speakable schema removes the guesswork. By marking specific sections, such as a summary paragraph, a key definition, or the first sentence of each section, you control which content AI systems prioritise.
Implementation
Implementing speakable schema involves adding a SpeakableSpecification to your page’s JSON-LD structured data. The cssSelector property accepts standard CSS selectors that point to the elements you want marked as speakable.
A common pattern is to mark the first paragraph of an article (which typically contains the summary or key point) and the paragraphs immediately following H2 headings (which often contain the most substantive content for each section).
Strategic Value
Speakable schema is one of the lower-effort, higher-signal structured data implementations. It requires minimal code, has no visible impact on your page, and directly tells AI systems which content to prioritise. For pages that target AI citations, particularly definition pages, FAQ content, and thought leadership, speakable schema is a worthwhile addition.
Best Practices
Mark content that is factual, concise, and self-contained. Avoid marking content that requires surrounding context to make sense. Each speakable section should be understandable on its own, because that is exactly how an AI voice assistant or citation system will use it.
Questions AI assistants answer about this topic
- How does speakable schema work?
- Speakable schema uses CSS selectors to point to specific parts of a page that are suitable for audio playback. For example, you might mark the first paragraph of an article and the content following each H2 heading. Voice assistants and AI models use these markers to select content that reads well aloud rather than attempting to read the entire page.
- Which AI models use speakable schema?
- Google Assistant and Google's AI features are the primary consumers of speakable schema. However, the markup also signals to other AI models which parts of your content are the most concise and quotable summaries, indirectly improving how all AI systems extract and cite your content.
- Should every page have speakable schema?
- Speakable schema is most valuable on content pages that contain clear, quotable information: blog posts, glossary entries, FAQ pages, and service descriptions. It is less useful on pages that are primarily navigational or transactional. Focus implementation on pages where you want AI models to extract and cite specific content.
Next Step
Want to know where your company stands?
We run 15-20 buyer queries across ChatGPT, Claude, Gemini, and Perplexity and show you exactly where you appear, and where you don't.
Get the Audit | from £750 ↗