- Speakable schema (schema.org/speakable) tells voice assistants exactly which sentences to read aloud — removing the guesswork from how your content sounds when spoken.
- Google's voice search uses speakable markup to source answers for Google Assistant responses, especially on news and informational content.
- Most small business websites have zero structured data, let alone speakable schema — early adoption is a real competitive edge.
- Speakable passages should be short (under 30 words), answer-first, and free of jargon or visual formatting that doesn't translate to audio.
- JSON-LD is the recommended implementation format — it lives in a script tag and doesn't require touching your visible HTML.
- Voice search queries skew conversational and local, making speakable schema especially valuable for businesses with a geographic service area.
The Problem Nobody Is Talking About
When someone asks their Google Home "What are the best local plumbers near me?" or "What time does [your business] open on Saturdays?", a voice assistant has to make a decision in milliseconds: which sentence on which page do I read aloud?
Without any guidance from the page itself, it guesses. It might pull the third paragraph. It might read a sentence that contains a hyperlink anchor text. It might pick something that sounds perfectly readable on a screen but is incomprehensible when spoken.
Speakable schema solves this. It's a piece of structured data markup — specifically schema.org/speakable — that you embed in your page's JSON-LD to explicitly tell voice assistants: these are the sentences worth reading aloud.
The concept sounds niche. It isn't. Voice queries now account for a substantial share of local search activity, and that share is growing as smart speakers and AI assistants become household fixtures. If your business depends on local visibility, speakable schema is one of the most underused tools available to you.
What Speakable Schema Actually Is
Speakable schema is a property within the schema.org/WebPage or schema.org/NewsArticle type that uses either CSS selectors or XPath expressions to point to specific sections of your page. When a voice assistant processes your page, it looks for this property and uses the identified passages as its preferred source material for spoken responses.
Here's a simple real-world example in JSON-LD:
{
"@context": "https://schema.org/",
"@type": "WebPage",
"name": "Emergency Plumber in Austin TX",
"speakable": {
"@type": "SpeakableSpecification",
"cssSelector": [".speakable-intro", ".speakable-hours"]
}
}
In this case, you're pointing at two CSS classes on your page — an introduction paragraph and your hours section — as the content Google Assistant should read aloud when your page is cited in a voice response.
The SpeakableSpecification type supports two targeting methods:
cssSelector— an array of CSS class or ID selectors (simpler, works for most CMS setups)xpath— XPath expressions for more granular targeting (better for complex page structures)
For most small business owners managing their own WordPress or Webflow site, CSS selectors are the right choice.
Why It Matters for Voice Search
Voice search is structurally different from typed search. When you type a query, you get a results page you can scan. When you speak a query, you get one answer read back to you. That single answer either comes from your business or it comes from your competitor.
The stakes are also higher in local contexts. Research on voice search behaviour consistently shows that voice queries are more likely to be local, transactional, and near-real-time in intent — exactly the queries small businesses need to win.
Google Assistant, in particular, uses speakable markup as a signal when generating voice responses from publisher content. According to Google's own developer documentation, pages with speakable markup are eligible to be featured in Google Assistant responses. Pages without it are not ineligible — but they're competing with pages that have given the algorithm a much clearer instruction set.
Think of it this way: structured data is a direct channel of communication between your content and search algorithms. Speakable schema is the part of that conversation that says, "here's what to say about us."
What Makes a Good Speakable Passage?
Not every sentence qualifies. Voice assistants have to read text aloud, which means passages need to work as spoken language — not as visual content.
Good speakable content:
- Answers a question directly in the first sentence
- Uses plain language, no jargon
- Is 20–30 words maximum per sentence
- Contains your business name, location, or service naturally
- Is factually stable (hours, services, expertise claims)
Bad speakable content:
- Bullet-point lists (assistants read these awkwardly)
- Sentences with embedded links or anchor text
- Paragraphs with technical terms, abbreviations, or product codes
- Content that references images ("as shown in the diagram above")
- Marketing superlatives that sound hollow when spoken aloud ("We are Austin's premier world-class provider of...")
The ideal speakable passage sounds like something a knowledgeable staff member would say if they picked up the phone. Concise, confident, and useful.
Speakable Schema in the Context of the Broader AEO Stack
Answer Engine Optimisation (AEO) is the discipline of making your content retrievable by AI-driven answer engines — including voice assistants, AI overviews, and LLM-based search tools. Speakable schema is one of its most specific tactical tools.
The broader AEO stack for a small business looks like this:
- FAQ schema — surfaces question-and-answer pairs in SERPs and AI answers
- HowTo schema — marks up step-by-step content for AI instructions
- LocalBusiness schema — verifies your name, address, hours, and service area
- Speakable schema — flags which sentences are voice-ready
Each layer adds precision. Each one reduces the chance that an algorithm misrepresents your business or simply skips you in favour of a competitor who has done the work.
The businesses winning in voice search right now are not doing anything magical. They are doing basic things completely — and speakable schema is one of those basic things that almost nobody has actually done.
The Implementation Reality
Here's the honest reality of implementing speakable schema on a small business website:
Time required: 20–45 minutes for someone comfortable with a CMS or basic HTML.
Skills required: Ability to add a JSON-LD script block to a page (either via a plugin, a theme's header/footer settings, or a tag manager). No coding background required.
Tools that help:
- Google's Rich Results Test validates your markup
- Schema Markup Validator checks for errors
- Yoast SEO, Rank Math, or similar WordPress plugins can simplify JSON-LD management
Current status of Google's rollout: Google initially launched speakable support for news publishers and has been gradually expanding it. The technical infrastructure exists in Google Assistant; adoption by non-news publishers is still relatively limited — which means acting now puts you ahead of the curve rather than catching up to it.
A Real Example: Local Service Business
Imagine you run an HVAC company in Phoenix. Your homepage has the following paragraph:
"Desert Air HVAC serves the Phoenix metro area with same-day emergency repairs, seasonal tune-ups, and full system installations. We're available 24/7 at (602) 555-0190."
That sentence is speakable gold. It contains your business name, service area, core services, availability, and phone number — in 29 words. Mark it up, point your speakable schema at it, and when someone asks Google Assistant "who does emergency HVAC repairs in Phoenix?", your page has handed the algorithm exactly what it needs.
Without the markup, the algorithm might pull your third pricing table row instead.
Common Mistakes to Avoid
Marking up too much content. Speakable passages should be the best 1–3 sentences on the page, not the entire page. Over-flagging reduces signal quality.
Targeting dynamic content. If your CSS selector points to content that changes based on user login state or JavaScript rendering, the markup won't resolve consistently for crawlers.
Ignoring mobile performance. Voice searches predominantly happen on mobile devices. If your page loads slowly on mobile, the best speakable markup in the world won't help — the assistant will find a faster page.
Setting it and forgetting it. When you update your hours, services, or location, update the speakable passages too. Stale spoken answers erode trust faster than stale written ones.
The Competitive Window Is Open Right Now
Adoption of speakable schema among small businesses is still negligible. Most SMB websites have no structured data at all — let alone this specific type. That's actually good news if you're reading this now. The competitive window is open.
Voice search is not a future trend. It is present tense. The businesses that mark up their content today are building a structural advantage that compounds over time — every new voice assistant interaction, every AI overview citation, every "near me" query is an opportunity to either show up or be invisible.
Twenty minutes of markup work is one of the highest-leverage SEO investments you can make this quarter. Do it before your competitors figure it out.
“Twenty minutes of markup work is one of the highest-leverage SEO investments you can make this quarter — do it before your competitors figure it out.”
| Area | Without Speakable Schema | With Speakable Schema |
|---|---|---|
| Passage selection | Voice assistant picks passages algorithmically — often poorly, pulling list items or link-heavy sentences | You explicitly designate the best 1–3 sentences, which are answer-first and spoken-language-ready |
| Brand accuracy in voice responses | Business name, hours, or service area may be missing from the spoken answer | Your flagged passage can include business name, location, and key services in a single spoken sentence |
| Implementation time | No action needed — but no competitive advantage either | 20–45 minutes of JSON-LD markup work delivers a durable structural advantage |
| Google Assistant eligibility | Page competes on generic algorithm signals with no explicit voice guidance | Page is eligible for direct Google Assistant citations per Google's developer documentation |
| Competitor differentiation | Virtually identical to every other small business site that hasn't implemented structured data | Ahead of the vast majority of SMB competitors who have never implemented speakable schema |
| Content quality control | Algorithm may read outdated, promotional, or visually-formatted text that sounds wrong spoken aloud | You control exactly what is spoken — and can update it whenever your services or hours change |
How to Implement Speakable Schema on Your Business Website
- 01Identify your best speakable passages. Find 1–3 sentences on your key pages (homepage, services page, contact page) that are under 30 words, answer-first, and contain your business name, location, or core service. Read them aloud — if they sound natural and complete, they're good candidates.
- 02Add CSS classes to your target HTML elements. In your CMS or page editor, wrap each target sentence or paragraph in a span or div and assign a distinct CSS class such as 'speakable-intro' or 'speakable-hours'. These classes won't affect your visual design — they're just hooks for your schema markup to reference.
- 03Write your JSON-LD speakable markup. Create a JSON-LD script block using the schema.org WebPage type with a speakable property that contains a SpeakableSpecification referencing your CSS selectors. Use Google's developer documentation as a template to ensure the syntax is correct.
- 04Add the JSON-LD block to your page. Paste the script block into your page's <head> section using your CMS header settings, a structured data plugin (Rank Math, Yoast SEO, Schema Pro), or Google Tag Manager. Make sure it fires on the correct pages and not site-wide.
- 05Validate your markup with Google's Rich Results Test. Visit search.google.com/test/rich-results, enter your page URL, and check that the speakable property is detected without errors. Also run it through schema.org's Schema Markup Validator to catch any type-level issues.
- 06Request re-indexing in Google Search Console. Open Google Search Console, go to the URL Inspection tool, enter your updated page URL, and click 'Request Indexing'. This accelerates Google's crawl of your new markup rather than waiting for the next natural crawl cycle.
- 07Monitor and update as your content changes. Set a quarterly reminder to review your speakable passages for accuracy — especially if your hours, services, or service area change. Stale spoken answers are worse than no answer because they actively mislead potential customers.