We Added Speakable Schema to 20 Pages

What is Speakable schema?

SpeakableSpecification is a schema.org type that marks sections of a page for reading aloud or quoting in conversational responses. It was designed for Google Assistant and voice search. It has found a second life in AI search. When ChatGPT, Gemini, or Perplexity need a concise, quotable answer, Speakable markup tells them where to look.

Adoption remains low. Our data from the Speakable adoption tracking page shows fewer than 10% of sites in competitive sectors use it. This is a genuine first-mover advantage for businesses willing to do the technical work.

What we implemented

We added SpeakableSpecification to 20 pages across this site, targeting two CSS selectors:

• .answer-capsule - the highlighted key takeaway boxes used throughout our content
• .key-takeaway - summary paragraphs designed to be directly quotable

The implementation uses JSON-LD within the page's schema markup. Here is the structure:

{
  "@type": "SpeakableSpecification",
  "cssSelector": [".answer-capsule", ".key-takeaway"]
}

This is embedded within the Article schema that already exists on each page. No separate script tag is needed. The Speakable specification simply becomes a property of the article's structured data.

What to realistically expect

We are not going to put a percentage on the uplift, because we do not have a controlled measurement and the credible public data on Speakable's effect on AI citations is thin. What we can say honestly: the markup helps machines locate your most quotable passages, it is cheap to add, and any benefit is strongest on pages with clear, concise answer paragraphs that AI models can lift directly. Treat it as one supporting signal, not a lever that moves results on its own.

Several factors influence whether it helps at all:

• Content quality matters more than markup. Speakable schema on poorly written content does not help. The markup highlights good content; it does not create it.
• Specificity of CSS selectors. Targeting precise, well-written sections performs better than applying Speakable to entire page bodies.
• Combination with other schema types. Pages that also include FAQ, Article, and Organization schema see compounding benefits from adding Speakable on top.

How to implement it yourself

The implementation is straightforward for any developer comfortable with JSON-LD:

1. Identify the CSS classes used on your most quotable content sections
2. Add a speakable property to your existing Article schema
3. Set the @type to SpeakableSpecification
4. List your target CSS selectors in the cssSelector array
5. Validate with Google's Rich Results Test

The whole process takes under 30 minutes for a developer who already has Article schema in place. For a deeper guide to structured data for AI search, see our schema markup guide.