Speakable specification
Also: SpeakableSpecification, speakable schema, voice search schema
Voice surfaces (Google Assistant, Siri, Alexa) and AI summarization (ChatGPT voice mode, Perplexity audio) preferentially extract content marked with SpeakableSpecification when generating spoken answers. Pages without Speakable schema rely on the model to guess which paragraph to read aloud, which produces inconsistent and often awkward voice output.
KailxLabs ships Speakable on every Article entity and on the homepage WebSite entity. The cssSelector typically targets the H1, the article lede, any paragraph carrying the answer-block class, and the TL;DR list. The hero answer capsule and the methodology Answer Capsules are the primary citation targets.
Speakable is not heavily weighted by ChatGPT or Gemini text retrieval, but it is a meaningful signal for voice-first surfaces that are an emerging share of total AI search traffic in 2026.
Cited facts
- Schema.org SpeakableSpecification uses cssSelector or xpath properties to identify quotable content.
- Voice surfaces represent an estimated 8 to 12 percent of total AI search query volume in 2026 and growing.