koira
speakable schemavoice searchstructured data

The Owner-Operator's Guide to Speakable Schema and Voice Search Visibility

KOIRA Team8 min read1,820 words
Speakable schema JSON-LD code block with voice search waveform and smart speaker icon
Intro
Breakdown
Solution
FAQ
◆ Key takeaways
  • Speakable schema uses CSS selectors or XPath expressions to point voice assistants at your most answer-ready sentences, rather than letting them guess.
  • Google's voice search and AI Overviews both draw from structured data — speakable markup gives your content a direct line to spoken results.
  • The best candidates for speakable markup are crisp, self-contained sentences: definitions, hours, FAQs, and short factual claims — not paragraphs of context.
  • Speakable schema is part of the broader Answer Engine Optimization (AEO) toolkit alongside FAQ, HowTo, and DefinedTerm schema types.
  • Owner-operators can implement speakable markup manually in JSON-LD without touching their site's template code — it lives in a single script block.
  • Voice search queries skew conversational and local — pages that answer 'near me' and 'what are your hours' questions in plain language benefit most.

What Speakable Schema Actually Is

When someone asks a voice assistant a question, the assistant has to pick a sentence — or a few sentences — to read aloud. It can't recite your whole page. So it makes a judgment call: which part of this content sounds most like a direct answer?

Speakable schema is how you take that judgment call away from the algorithm and make it yourself.

Formally, speakable is a property on the Schema.org Article type that accepts either CSS selectors or XPath expressions pointing to specific sections of your page. When a voice assistant or AI search engine encounters that markup, it knows: these are the sentences the publisher considers most suitable for audio playback.

Google introduced support for speakable in 2018 for news publishers, but its relevance has expanded considerably as AI-generated answers — in Google's AI Overviews, Perplexity, and smart-speaker responses — increasingly draw from structured data rather than raw HTML parsing.

Why It Matters More Now Than It Did in 2018

In 2018, speakable schema was mostly a news-publisher tool. Google used it to feed content to Google Assistant on smart speakers. Most small-business owners had no reason to care.

That's changed for a few reasons:

Voice search volume has matured. Roughly 27% of the global online population uses voice search on mobile, and smart-speaker households routinely ask local questions — hours, directions, availability, pricing — that your business page should be answering.

AI search engines parse structured data first. Systems like Google's AI Overviews and Perplexity don't just scrape text; they weight content that has been explicitly labeled. A page with speakable markup signals to those systems: this section was written to be a direct answer. That signal matters.

The gap between a good page and a spoken answer has widened. You can have excellent content that ranks well in blue-link results and still get skipped entirely in voice results, because voice engines need a clean, self-contained sentence — not a well-structured paragraph. Speakable markup bridges that gap by pre-selecting your best sentences.

As the broader shift toward AI-driven search answers continues, the pages that have done the work of labeling their most answer-ready content will have a structural advantage over those that haven't.

What Content Should Be Marked Speakable

Not everything on your page deserves to be read aloud. Voice assistants have roughly 20–30 seconds of listener patience before attention drops. That means your speakable-tagged content needs to be short, self-contained, and factually complete.

Good candidates:

  • Definitions and direct answers — "Our cancellation policy requires 24 hours' notice."
  • Business hours and location details — "We're open Monday through Saturday, 9 a.m. to 6 p.m."
  • FAQ-style responses — The answer sentence in a Q&A block, not the surrounding context.
  • Key claims or summaries — The first sentence of a section that could stand alone as an answer.
  • Pricing or service details — "Haircuts start at $35 for adults."

Bad candidates:

  • Long paragraphs that require context to make sense
  • Sentences with pronouns that reference earlier content ("It works by...")
  • Calls to action ("Click here to book")
  • Navigation text, headers, or boilerplate

The test: if someone heard only that sentence with no context, would it answer a reasonable question completely? If yes, it's speakable. If it needs setup, it isn't.

The Technical Mechanics

Speakable schema is implemented in JSON-LD, the same format used for FAQ, HowTo, and LocalBusiness schema. It lives in a <script type="application/ld+json"> block in your page's <head> or just before </body>.

Here's the basic structure:

{
  "@context": "https://schema.org",
  "@type": "Article",
  "name": "Your Page Title",
  "speakable": {
    "@type": "SpeakableSpecification",
    "cssSelector": [".speakable-intro", ".faq-answer", "#hours-summary"]
  },
  "url": "https://yourdomain.com/your-page"
}

The cssSelector array points to elements on your page by their CSS class or ID. When Google's crawler or a voice assistant hits your page, it looks for those elements and treats their text content as your speakable passages.

Alternatively, you can use xpath instead of cssSelector if your CMS generates predictable XPath structures — but CSS selectors are easier to maintain for most owner-operators.

One important note: your CSS selectors have to actually match elements that exist on the rendered page. If you tag .faq-answer but your CMS renders FAQ answers as .faq__body, the markup will point at nothing. Always verify with Google's Rich Results Test after implementation.

Speakable Schema vs. Other Schema Types

Speakable doesn't replace your other structured data — it works alongside it. Here's how the types relate:

  • FAQPage schema tells search engines you have a Q&A block and surfaces the Q&A in rich results. Speakable tells voice assistants which of those answers to read aloud.
  • LocalBusiness schema provides your NAP (name, address, phone), hours, and service area. Speakable complements it by flagging the human-readable summary of that information.
  • HowTo schema structures step-by-step content. Speakable can tag the intro sentence of each step if it reads cleanly in isolation.
  • DefinedTerm schema labels definitions. If you have a definition that also qualifies as speakable, you can implement both.

Think of speakable as the audio layer on top of your existing structured data strategy. It doesn't replace any of the other types — it adds a voice-specific signal that the others don't provide.

For a broader look at how website content structure affects AI search engines, the principles overlap: both reward content that is organized, labeled, and self-contained at the sentence level.

How Speakable Fits Into AEO (Answer Engine Optimization)

Answer Engine Optimization is the practice of structuring content so that AI systems — not just human readers — can extract and surface accurate answers. Speakable schema is one of the most direct AEO tools available because it's the only schema type that explicitly addresses the audio output use case.

Most AEO work focuses on making content machine-readable: clear headers, short paragraphs, FAQ blocks, schema markup. Speakable takes that one step further by saying: among all the machine-readable content on this page, these specific sentences are optimized for being spoken aloud.

For local businesses, this is particularly high-value. Voice queries skew heavily local and transactional — "what time does [business] close," "does [business] take walk-ins," "how much does [service] cost." These are exactly the kinds of short, factual answers that speakable markup is designed to surface.

Voice engines don't read your page — they pick a sentence. Speakable schema lets you choose which one.

Common Implementation Mistakes

Tagging too much. If you mark 15 different elements as speakable, you've given the voice assistant no real guidance. Prioritize 2–4 elements per page.

Using selectors that don't exist. Always verify your CSS selectors against the actual rendered HTML. CMS themes often add prefixes or modify class names.

Tagging content that requires context. "As mentioned above, this approach works best when..." is not speakable. It needs context to make sense.

Forgetting to include the url property. Google's implementation requires the url property on the Article object to associate the speakable content with the correct page.

Not testing after CMS updates. If your theme updates or your CMS changes how it renders elements, your selectors may break silently. Schedule a quarterly check with the Rich Results Test.

Who Should Prioritize This

Speakable schema delivers the most value for:

  • Local service businesses (salons, restaurants, clinics, auto shops) where customers ask voice assistants about hours, location, and availability
  • Content publishers with FAQ-heavy or definition-heavy pages
  • E-commerce stores with product pages that answer common questions (sizing, materials, shipping times)
  • Professional services (lawyers, accountants, consultants) where clients ask qualifying questions before calling

If your site already has structured data for local business and you're investing in AEO, speakable is the logical next layer. It takes less than an hour to implement on a page and doesn't require any changes to your visible content or site design.

Voice engines don't read your page — they pick a sentence. Speakable schema lets you choose which one.

Save this for later
Get a PDF copy of this post →
Drop your email, we’ll send you the full piece as a clean PDF. Plus the weekly KOIRA roundup.
Title: Speakable Schema: What It Is and Why Voice Search Needs It
Speakable Schema
A Schema.org structured-data property that uses CSS selectors or XPath expressions to identify specific page sections as suitable for audio playback by voice assistants and AI search engines.
SpeakableSpecification
The Schema.org object type used inside the speakable property to define which page elements — via cssSelector or xpath — contain voice-ready content.
Answer Engine Optimization (AEO)
The practice of structuring and labeling web content so AI-driven answer engines can accurately extract and surface it as a direct response to user queries.
CSS Selector (in schema context)
A pattern string used in speakable schema to target specific HTML elements by class, ID, or tag so voice assistants know exactly which text to retrieve.
Voice Search
A search modality where users speak queries to a device and receive a spoken answer, requiring search engines to select a single short passage rather than return a list of links.
Speakable Schema: Without It vs. With It
AreaWithout speakable markupWith speakable markup
Voice assistant answer selectionAlgorithm guesses which sentence to read — often picks boilerplate or navigation textYou pre-select 2–4 answer-ready sentences; assistant reads those
AI Overview inclusionContent competes on text quality alone with no explicit audio signalStructured speakable passages are weighted as explicitly answer-ready content
Local query responsesHours, pricing, and policy details buried in paragraphs get skippedTagged summary sentences surface directly in spoken local search results
Implementation effortNone required — but no control over what gets read aloudOne JSON-LD block per page, under an hour to implement and test
Content quality incentiveNo structural reason to write self-contained answer sentencesForces you to write crisp, context-free answers that benefit all channels
Maintenance overheadNothing to maintain — but nothing working eitherQuarterly selector check after CMS updates; otherwise set-and-forget

How to Implement Speakable Schema on Your Website

  1. 01
    Identify your best answer-ready sentences. Go through your page and find 2–4 sentences that could answer a spoken question completely without any surrounding context. Good targets: your hours summary, a one-sentence policy statement, the direct answer in a FAQ block, or a pricing summary.
  2. 02
    Assign CSS classes or IDs to those elements. In your CMS or HTML, add a specific class (e.g., `speakable-answer`) to the elements containing those sentences. If they already have unique classes or IDs, note those — you'll reference them in your schema.
  3. 03
    Write the JSON-LD block. Create a `<script type="application/ld+json">` block with `@type: Article`, a `speakable` property containing a `SpeakableSpecification` with your CSS selectors in an array, and the canonical `url` of the page. Paste the block into your page's `<head>` section.
  4. 04
    Verify selectors match your rendered HTML. Open your browser's developer tools, run `document.querySelectorAll('.your-selector')` in the console, and confirm it returns the correct elements. A selector that matches nothing in the live DOM will silently fail.
  5. 05
    Validate with Google's Rich Results Test. Paste your page URL (or the raw HTML) into search.google.com/test/rich-results and check that the speakable property is detected and the extracted text matches your intended passages. Fix any errors flagged before moving on.
  6. 06
    Request indexing in Google Search Console. In Google Search Console, use the URL Inspection tool to request a re-crawl of the updated page. This speeds up the time before Google processes your new structured data rather than waiting for the next natural crawl.
  7. 07
    Schedule a quarterly selector audit. Add a recurring calendar reminder to re-run the Rich Results Test on your speakable pages after any CMS theme update or major content revision. Class names and element structures change silently during updates and can break your selectors without any visible warning.
FAQ
Does Google officially support speakable schema for non-news sites?
Google's original speakable documentation was targeted at news publishers, but the schema.org property itself is not restricted to news content, and Google's voice search systems read structured data from all page types. Many SEO practitioners implement speakable on local business, FAQ, and e-commerce pages with measurable results. Google's Rich Results Test will validate speakable markup regardless of site category, which is a practical signal that it processes the property broadly.
How is speakable schema different from FAQ schema?
FAQ schema (FAQPage type) structures question-and-answer pairs so they can appear as expandable rich results in Google Search. Speakable schema doesn't create any visible search result feature — instead, it flags specific text passages as suitable for audio playback by voice assistants. The two can work together: FAQ schema surfaces your Q&A in visual results, while speakable markup tells voice engines which answer sentences to read aloud.
Will adding speakable schema hurt my regular search rankings?
No. Structured data is a supplementary signal — it doesn't replace or compete with your existing ranking factors. Adding speakable markup correctly will not negatively affect your rankings in standard text-based search. The only risk is incorrect implementation (e.g., selectors that point to non-existent elements), which Google will simply ignore rather than penalize.
How do I know if my speakable markup is working?
Use Google's Rich Results Test (search.google.com/test/rich-results) to validate that Google can parse your speakable JSON-LD and that the CSS selectors resolve to actual content on your page. The tool will show you which text passages the markup points to. Beyond validation, you can monitor voice search performance indirectly through Google Search Console's query data — look for question-format queries where your page appears and track click-through rates over time.
How many elements should I mark as speakable on a single page?
Two to four elements per page is the practical sweet spot. Marking too few gives voice assistants little to work with; marking too many dilutes the signal and risks including passages that aren't truly self-contained. Prioritize your page's single best direct-answer sentence first, then add one or two supporting passages — such as a key FAQ answer or a summary of your hours or pricing.
Does speakable schema work with Alexa and Siri, or only Google Assistant?
Google has the most documented support for speakable schema, particularly through Google Assistant and AI Overviews. Alexa and Siri use different indexing pipelines and don't publicly document speakable support, but they do parse structured data from web pages as part of their knowledge retrieval. Implementing speakable is primarily a Google optimization today, though the clean, self-contained content it encourages you to write benefits all AI-driven answer engines.
Find KOIRA on
XLinkedInFacebookCrunchbaseWellfoundF6S
Keep reading
Data
What Share of Searches Now End in AI Answers?
7 min read
Guides
SEO vs GEO vs AEO: What Each One Does and Why You Need All Three
9 min read
Guides
How to Structure Website Content for AI Search Engines
9 min read
Guides
The 5-Touch Sales Sequence Every Small Business Should Automate
9 min read
Stay in the loop
New posts, straight to your inbox.
Marketing and sales insights from the KOIRA team. No filler.
Speakable Schema: What It Is and Why Voice Search Needs It
Get KOIRA