Speakable Schema and the Return of Voice Search
Semantic Summary
Idea: Speakable schema marks the sections of an article or page that suit text-to-speech playback. Google documents it as a beta feature for news publishers, so most sites gain nothing from the markup itself.
Challenge: Guides promise voice and AI visibility from a few lines of code. Google’s documentation describes a narrow, long-running beta, and no one has published how AI assistants use the markup.
Summary: Check eligibility first. Then write short, self-contained sections that sound right read aloud, and treat the markup as optional.
Related reads: Voice Search Optimization in 2026 · Multimodal SEO: Optimizing for Text, Image, and Voice Search · The FAQ Strategy for Voice Queries
Speakable schema is structured data that marks the sections of an article or web page best suited to text-to-speech playback. Google documents it as a beta feature, mainly for news publishers, and describes it for English content in the U.S. on Google Home devices. For other sites, the markup is unlikely to change anything on its own. What helps everywhere is writing short, clear answers that sound right when read aloud. This guide is marketing advice, not a guarantee of results.
What is speakable schema?
Speakable is a property from schema.org. It points to parts of a page that are especially suited to being spoken. It uses the speakable property with a SpeakableSpecification, and it can apply to an Article or a WebPage.
The markup selects text. It does not create audio, and it does not make a page rank. A device or app that supports it can read the marked sections aloud with text-to-speech.
What does Google say about speakable markup?
Google’s Speakable documentation labels the feature as beta and subject to change. As described there and in reports:
| Topic | What the documentation says |
| Purpose | Google Assistant uses it to answer topical news queries on smart speakers |
| Who is eligible | News publishers, through Google’s Publisher Center |
| Where it works | English content in the U.S. on Google Home devices; other countries and languages are hoped for |
| Section length | About 20 to 30 seconds, or two to three sentences |
| What to avoid | Datelines, photo captions, and source attributions |
| Types | Article and WebPage |
Because of those limits, most sites cannot use the Google feature at all. Check the current page before you invest time.
Does speakable schema affect rankings or voice search visibility?
There is no documented ranking effect. Some third-party reports say speakable has essentially no effect for sites outside news. Treat those reports as secondary, because Google’s page is the source of truth.
| Claim | Evidence | How to treat it |
| Speakable can help eligible news content play on supported speakers | Documented by Google (beta) | Use it if you are eligible |
| Speakable improves rankings | Not documented | Do not expect it |
| Speakable gets non-news pages read aloud | Not documented; reports say no | Do not rely on it |
| ChatGPT or Perplexity use speakable markup | Not documented | Treat as unproven |
| Short, clear answers help voice and AI tools | Plausible, since both extract short passages | Prioritize this |
Our guide to AI visibility experiment design shows how to test one change at a time.
Is voice search making a return?
No one has published a reliable measure of a “return”. What is clear is that voice assistants and chatbots with voice modes keep changing, and the way they choose answers is often not documented.
So avoid plans that depend on a voice trend you cannot measure. Instead, write content that works in any channel where an answer is read or spoken. Our guides on voice search optimization and multimodal SEO cover the wider context. We do not repeat their figures here, because they are not verified.
What should you mark as speakable?
Mark short sections that make sense without the rest of the page and that sound clear when spoken.
- The headline and a summary. They give the story in a sentence or two.
- A direct answer. One self-contained answer to the main question.
- Key facts in plain sentences. Avoid dense lists and tables.
Do not mark datelines, captions, source lines, or text full of abbreviations, symbols, and numbers. Ask one test question: would you be comfortable with this being read aloud and attributed to you?
How do you write content that sounds right read aloud?
This advice works with or without markup.
- Lead with the answer. Put it in the first sentence.
- Keep sections short. Aim for two or three sentences.
- Use plain words. Avoid jargon and unexplained abbreviations.
- Name the subject. Do not rely on “it” or “this” across sentences.
- Spell out numbers and units. Read the sentence aloud to check it.
- Test it. Use a text-to-speech tool or read the text aloud yourself.
Our Atomic Answer Framework describes the same pattern for AI answers.
How NEURONwriter helps: for a voice-style question, NEURONwriter SERP analysis and People Also Ask list show how top pages phrase and answer related questions. Neuro, the inline assistant, can run short commands such as shortening a paragraph or rewriting it in a few plain sentences. The content score measures topical coverage, not how a text sounds, so read it aloud yourself.
What does speakable JSON-LD look like?
Here is a simplified example with placeholder values. It uses CSS selectors to point to sections of the page.
<script type=”application/ld+json”>
{
“@context”: “https://schema.org”,
“@type”: “Article”,
“name”: “Example news story”,
“url”: “https://example.com/news/example-story/”,
“speakable”: {
“@type”: “SpeakableSpecification”,
“cssSelector”: [“#headline”, “.summary”]
}
}
</script>
The selectors must match real elements on the page. Schema.org also allows XPath, and it allows id-value references. Google’s page describes the options it supports.
How do you add and test speakable markup?
Follow these steps if you are eligible.
- Confirm eligibility. Read the current documentation and the Publisher Center requirements.
- Choose sections. Pick a headline and a short summary.
- Add the markup. Use JSON-LD with cssSelector or xpath.
- Check the selectors. Make sure each one returns the intended text.
- Validate. Use the Schema Markup Validator for syntax, and review Search Console for structured data errors.
- Review after changes. Template updates can break selectors.
How NEURONwriter helps: NEURONwriter Generate Schema Markup option creates JSON-LD from page content, but its FAQ names Blog, HowTo, Review, Product, and FAQ as examples and does not list Speakable. Check the type list in your editor. If Speakable is not offered, add it with your theme, a plugin, or hand-written JSON-LD. The AI Readiness audit can still check whether structured data is present on a page.
What is the difference between speakable and FAQ schema?
They solve different problems.
- Speakable. Marks sections of a page for audio playback.
- FAQ schema. Labels question-and-answer pairs. Google stopped showing FAQ rich results on 7 May 2026, although the type is still valid. See our FAQ schema guide.
Neither one is a ranking switch. Use either only when it describes visible content.
How do you check what voice and AI tools say about you?
Test the outcome, because you cannot see which signals each tool uses.
- Write spoken-style questions. Use the way people ask, such as “what is [topic]” or “how do I [task]”.
- Ask across tools. Try the assistants and chatbots your audience uses, in text and in voice mode where available.
- Log the answers. Note the wording, the sources, and any errors.
- Fix the source page. Update the page that holds the wrong or unclear answer.
Our citation-checking checklist explains the method.
How NEURONwriter helps: the AI Visibility module monitors the questions you add across Google AI Overviews, Google AI Mode, ChatGPT, and Perplexity, with a visibility score from 0 to 100 for each. It tracks text answers from those four engines, not voice assistants, so test voice tools yourself.
What are the basics of speakable schema at a glance?
This short guide collects what to remember about speakable schema markup.
- Use speakable schema only if you are eligible. Google describes it as a beta for news publishers.
- Mark sections of an article or webpage. Keep them to two or three sentences.
- Use JSON-LD where you can. Implement it with cssSelector or xpath.
- Avoid datelines and captions. They sound confusing when read aloud.
- Do not expect SEO gains. Speakable is not documented as a ranking factor.
- Write for the ear. Short, clear answers help in every channel.
Speakable markup supports audio playback of content for supported devices. It does not replace good writing or a clear page structure.
What should you know about voice assistants and search results?
Voice assistants such as Siri and Alexa choose answers from their own sources, and speakable markup is not documented for them. A response spoken by a device may come from a featured result, a knowledge source, a news feed, or a language model. No one outside the platform can say which, so do not assume that a single tag controls it.
The practical advice is to optimize for the query, not the device. Write precise answers to the questions your audience asks, give each answer its own specific section, and keep the page informational and easy to scan. When someone interacts with a speaker, a phone, or a chatbot, the same clear sentence can be read, spoken, or cited.
- Mobile and speakers. Many spoken questions come from a mobile device or a speaker, so short answers matter more than long ones.
- News articles. The documented speakable feature applies to news articles from eligible publishers. Service pages and product pages are not part of that documentation.
- Search rankings. Speakable is not a way to move search rankings. It only marks sections for audio playback using text-to-speech.
- Implementation. If you follow the guideline, implementing structured data for speakable takes a few lines, and the markup allows search engines and devices to find the right text.
- Testing. The Rich Results Test does not cover every beta feature, so check the current documentation for the right tool.
Content for Google, for a speaker, or for an AI chatbot should meet the same standard: be accurate, be short where possible, and do not cite facts you cannot support. Appropriate text-to-speech conversion depends on the device, so a TTS preview on your own devices is the best check. If you ever need a deeper answer, link to the full page for the reader.
Common speakable schema mistakes
- Expecting rankings. No documentation says speakable improves them.
- Marking long or complex text. It sounds poor read aloud.
- Marking datelines and captions. Google advises against it.
- Selectors that match nothing. Check that each one finds real text.
- Assuming it works for any site. Eligibility is narrow.
- Skipping the writing. Markup cannot fix an unclear answer.
How does NEURONwriter help with voice-ready content?
NEURONwriter helps you plan clear answers, tighten them, and watch how AI describes your content. It does not generate speakable markup unless your editor’s type list shows it, and it does not test voice assistants.
- Plan the questions. SERP analysis and People Also Ask show how top pages phrase related questions. The content editor and score show whether a draft covers what top pages cover.
- Write short answers. Neuro runs commands such as shortening or rewriting a paragraph, and Content Designer builds an outline with headings.
- Check presence. AI Readiness checks whether structured data and AI crawler access are in place.
- Monitor AI answers. AI Visibility tracks Monitored Discussions across four AI engines, with visibility scores and sentiment.
You can run a keyword analysis in NEURONwriter before you write your next answer page. Treat every score as a guide, because no tool can guarantee rich results, voice answers, or citations.
FAQ
What is speakable schema, and how does it work?
Speakable schema marks sections of an article or page that are best suited to text-to-speech playback. A supporting device or app can read those sections aloud.
How do you implement speakable schema on a webpage?
Add JSON-LD with a speakable SpeakableSpecification that points to page sections by cssSelector or xpath. Check that the selectors match real text, and validate the syntax.
What content should you mark as speakable?
Short, self-contained sections such as the headline and a brief summary. Avoid datelines, captions, source lines, and text full of symbols.
Does speakable schema affect Google rankings?
Google does not document a ranking effect. Treat it as a feature for audio playback, not an SEO lever.
Is speakable a ranking factor?
No documentation says so. Do not add it for ranking reasons.
Does speakable work outside the US?
Google describes it for English content in the U.S. and hopes to expand it. Check the current documentation for any change.
Can I use speakable schema on pages that are not articles?
The property can apply to Article and WebPage, but Google’s feature is described for news content. For other pages, there is no documented Google feature.
How many sections should I mark as speakable on a single page?
Google gives no fixed number. It recommends sections of about 20 to 30 seconds, so keep the markup to a few short sections.
Should I mark up the whole article?
No. Mark only short sections that make sense read aloud, such as the headline and a summary.
What is the difference between speakable schema and FAQ schema?
Speakable marks sections for audio playback. FAQ schema labels question-and-answer pairs, and Google no longer shows FAQ rich results.
Is speakable schema supported by ChatGPT and Perplexity?
No support is documented. Treat any claim about it as unproven, and test your own prompts.
Does speakable cover writing as well as speaking?
No. It is about audio playback of text through text-to-speech, not about writing.
Why is schema markup important for voice search?
It can describe a page to machines, but voice answers depend on many factors. Clear, short, accurate answers matter more than any single tag.
How can I test and validate schema markup?
Use a schema validator for syntax, check that your selectors match real text, and review Search Console for structured data errors.
What are the best practices for schema markup?
Match the visible page, keep one source of markup, use accurate values, validate it, and do not rely on markup alone. Follow Google’s documentation for each type.
Does FAQ schema still work?
The type is still valid, but Google stopped showing FAQ rich results on 7 May 2026. It remains an optional description of visible content.
