Speakable Schema (Healthcare) — the markup that makes a page voice-answerable.
Speakable Schema (Healthcare) is a WebPage + SpeakableSpecification JSON-LD block that marks specific CSS selectors on a page as suitable for voice-assistant / AI-Overview reading. Applied by ICG to money pages + FAQ H2 answers on 320+ posts.
How it works.
Speakable is a structural signal, not a content signal. When Google Assistant or an AI Overview extractor decides your page is the answer to a spoken query, the crawler looks for a WebPage schema block containing a SpeakableSpecification. Inside that spec, you nominate one or more CSS selectors — the exact HTML nodes the machine is allowed to read aloud. Without Speakable, the extractor guesses. It might pull a footer label, a menu item, or a compliance disclaimer as the "answer." With Speakable, the sentence you engineered — a 40-70 word plain-language reply — is what the patient or buyer hears. ICG treats every FAQ H2 answer as speakable-eligible, provided it passes the compliance gate.
- WebPage wrapper. A single WebPage JSON-LD block that names the page, its URL, and its primary language (typically en-IN for ICG's Indian healthcare corpus).
- SpeakableSpecification block. Nested inside WebPage, an array of cssSelector strings pointing to the exact nodes eligible for voice-read.
- Selector hygiene. Every selector is class-scoped (never a raw tag) so a template refactor does not silently break the speakable contract.
- Compliance gate. Every speakable node is auto-audited against NMC Code of Ethics, ASCI, and DPDP filters before deploy — no promise-language selector ships.
- FAQPage pairing. Speakable rides alongside FAQPage schema; the FAQ Q builds the rich-result carousel, the FAQ A becomes the voice answer.
- Weekly extraction audit. Search Console AI-Overview appearance queries are reconciled against the speakable corpus to confirm the right sentences are being lifted.
Why it matters for
Indian healthcare.
Speakable multiplies whatever sentence you nominate — the risk and the reward both scale. A voice answer that promises "guaranteed cure" or "100% success rate" violates the NMC Code of Ethics and ASCI unsubstantiable-superlative rules the moment Assistant reads it aloud, and the violation is now audible instead of buried in a schema payload. For fertility pages, an ART Act 2021 success-rate claim spoken by Assistant is a public act, not a footnote. For sex-determination-adjacent gynaecology pages, PC-PNDT language must never be nominated as speakable. DPDP Act 2023 Section 8 blocks any selector that could read back a patient testimonial as a spoken "answer." ICG's speakable audit runs every selector through these filters before the JSON-LD merges to production — a clean speakable corpus is also a compliant one.
How ICG uses this.
Speakable Schema is a default deploy target on YODA — ICG's AI-native healthcare YouTube + AIO platform. Across 320+ posts in the current corpus, Speakable rides alongside FAQPage schema on every clinical money page and every 6+ H2 article. The FAQ H2 answer is the voice payload; the definition paragraph directly under H1 is the AIO snippet payload. Weekly Search Console reconciliation confirms which selectors are actually being lifted into AI Overviews, and the ones that are get compliance-hardened first. 172 posts carrying FAQPage schema after Turn 16 all now carry paired Speakable selectors. The three ranking races — YouTube search, Google web, AI Overview — all three read structural signals; Speakable is the one that governs what a machine speaks.
| Selector target | Speakable payload | Compliance filter |
|---|---|---|
| H1 definition para | 40-70 word plain-language definition of the term or service | NMC · ASCI superlatives |
| FAQ H2 questions | The literal spoken query patients ask | PC-PNDT phrasing filter |
| FAQ first-answer sentence | The voice-eligible answer — 15-30 words, no promise language | ART Act success-rate filter |
| Money-page service line | One-line service description above the CTA | ASCI + DCI (for dental) |
| Compliance footer | Never nominated as speakable | Auto-excluded |
| Patient testimonials | Never nominated as speakable | DPDP Section 8 |
Frequently asked questions.
Does Speakable Schema actually get read by Google?
Speakable is an active read for Google Assistant news-briefing surfaces and a strong hint for AI Overview snippet extraction. It does not guarantee inclusion. What it does guarantee is that when Google decides to speak or cite a page, the selector you nominated is the sentence it lifts — instead of a stray navigation label or a compliance footer.
Which CSS selectors should a healthcare page mark speakable?
The definition paragraph directly under the H1, the H2 questions in the FAQ block, and the first answer sentence under each H2. On money pages, the one-line service description. Avoid marking the compliance line, the doctor-registration block, or the medico-legal disclaimer — those must never be spoken as the answer to a clinical query.
How is Speakable different from FAQPage schema?
FAQPage tells Google the page contains Q/A pairs so it can build a rich-result carousel. Speakable tells Google which sentences on the page are safe to read aloud in a voice answer. The two work together — FAQPage builds the visible-search snippet, Speakable guides the AI Overview and Assistant voice read. ICG ships both on every clinical money page.
Does Speakable violate NMC or DPDP compliance?
Not on its own, but a careless selector can. If Speakable marks a sentence that promises cure rates, names an individual doctor with a claim, or reads back a patient testimonial, the voice answer inherits that violation. ICG audits every Speakable selector against NMC Code of Ethics and ASCI unsubstantiable-superlative filters before the JSON-LD ships to production.
Read next.
Get your speakable payload
audited and shipped.
ICG runs a free 30-min Brand & Growth Diagnostic that audits your page schema against Speakable + FAQPage + VideoObject standards — and shows which selectors are safe to nominate under NMC, ASCI, DPDP, and ART Act.
