If you run a B2B services shop, you probably judge ideas by what they move, not how shiny they look. Here is the blunt version: speakable has a narrow sweet spot. It tells voice assistants which two or three lines in a story to read out loud. If you publish real news and want a little extra audio reach, it is worth a small test. If your site is mostly service pages and evergreen posts, park it near the bottom of the backlog.
speakable schema markup
Speakable markup is a pointer. You label two or three short sentences on a page, and a voice assistant can read those lines when someone asks for news. The intent is hands-free listening - useful when people cannot scan a page but still want the gist.
Decision snapshot for busy B2B leaders: only lean in if you publish news or frequent updates. Speakable showed up first in Google Assistant news contexts, not standard web results, so do not expect a shiny new rich result in traditional Search. It is still treated as limited or beta in various references. Schema.org defines the property, and Google’s guidelines in Search Essentials and the broader context in How Search works explain how content is evaluated. For pending status, see Speakable.
- It works best on Article or NewsArticle pages with a clean, two-sentence summary.
- Even perfect code does not guarantee a readout. Selection is algorithmic and competitive.

Country and language availability
Google’s rollout focused on English content in the United States via Google Home and Assistant. As of now, coverage is still limited. Before you spend engineering cycles, review current documentation to confirm availability and any regional or language support.
Even with clean implementation, there is no promise your story gets picked. Assistant experiences sample a small set of articles; some good pieces never make the cut. If your audience sits outside supported locales or languages, consider deferring the work or investing in alternatives that actually reach your market.
speakable schema for publishers
Who benefits most? Publishers - sites shipping timely stories with recognizable headlines. The property appears on Article, NewsArticle, and sometimes WebPage. It is not meant for service landers, gated collateral, or product copy.
Editorial guidance that keeps things smooth:
- Mark up two or three tight sentences - usually the lede or a crisp summary.
- Do not tag navigation, ads, bylines, datelines, or legalese.
- Keep the selected text stable in the DOM. If the target moves, the readout breaks.
Schema.org lists speakable on Article and WebPage (NewsArticle inherits from Article). In practice, structure the headline and summary for out-loud clarity and place them in elements with stable IDs or classes. Fragile selectors are the fastest way to sabotage yourself.
json-ld speakable examples
Here are JSON-LD examples aligned with Schema.org and Google’s guidance. If you need a refresher on the format, see JSON-LD.
NewsArticle using CSS selectors:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "NewsArticle",
"headline": "Northwind Cloud Releases 2025 SOC 2 Type II Report",
"datePublished": "2025-08-01T09:00:00Z",
"dateModified": "2025-08-01T09:00:00Z",
"url": "https://example.com/news/northwind-soc2-2025",
"speakable": {
"@type": "SpeakableSpecification",
"cssSelector": [
"#story-headline",
".story-summary"
]
},
"author": {
"@type": "Organization",
"name": "Northwind Cloud"
},
"publisher": {
"@type": "Organization",
"name": "Northwind Cloud",
"logo": {
"@type": "ImageObject",
"url": "https://example.com/static/logo.png"
}
}
}
</script>
NewsArticle using xpath:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "NewsArticle",
"headline": "AI Infrastructure Spending Hits New Quarterly High",
"url": "https://example.com/news/ai-infra-spend-q3",
"speakable": {
"@type": "SpeakableSpecification",
"xpath": [
"/html/body/main/article/header/h1",
"/html/body/main/article/section[@class='summary']/p[1]"
]
}
}
</script>
WebPage example for a briefing hub with short summaries:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "WebPage",
"name": "Weekly Tech Briefing",
"url": "https://example.com/insights/weekly-briefing",
"speakable": {
"@type": "SpeakableSpecification",
"cssSelector": [
".briefing-headline",
".briefing-summary"
]
}
}
</script>
Details that actually matter:
- Keep the spoken chunk brief - roughly 50 to 80 words (20 to 30 seconds when read).
- Selectors must resolve to visible text at render time. Hidden or late-loaded summaries often yield nothing.
Testing pointers:
- Validate with the Schema Markup Validator and confirm the speakable property resolves to the intended nodes.
- Use the Rich Results Test to check render and parsing behavior.
- In DevTools, sanity-check selectors:
document.querySelectorAll(".story-summary")
should return the right element(s). - Once live, request indexing in Search Console and give it time to re-crawl.
schema.org speakable property
The speakable property marks the parts of a page that work best for text-to-speech. It accepts either a SpeakableSpecification (recommended) or a URL. SpeakableSpecification supports content locators like cssSelector
or xpath
. For Google’s implementation, pick one approach per page - do not mix cssSelector
and xpath
.
Valid on Article and WebPage. NewsArticle works by inheritance. See Article and WebPage for definitions, and the property details at schema.org/speakable. You can target multiple on-page elements - for example, headline plus summary - by listing more than one selector in the same SpeakableSpecification. It is a highlight, not a transcript.
One common snag: your speakable targets are separate from the headline or description fields in your JSON-LD. You still need a proper headline on the article, and your speakable selectors should point to on-page elements that match the summary a user will actually hear.
implement speakable markup for news articles
A simple, durable workflow:
- Pick the right pages: start with news items that already get attention via Search or Google News.
- Choose the audio-friendly text: the lede or a plain-English two-to-three-sentence summary - no insider jargon.
- Add JSON-LD with SpeakableSpecification: use stable IDs or classes so selectors survive design tweaks.
- Validate: no schema errors, and the selectors match visible DOM text.
- Request indexing and wait for a fresh crawl.
- Monitor: watch impressions and clicks from news surfaces where available, and check Search Console for structured data issues.
CMS and implementation notes:
- Lock down selector stability. Coordinate with your theme or CMS team; a casual class rename can break the readout.
- Avoid duplicate JSON-LD blocks for the same entity. One clear script per page is easier to debug.
- Get canonicalization right. The canonical should be the indexed, publicly reachable URL.
- Heavy client-side rendering? Ensure the summary exists in the rendered DOM that Googlebot sees.
Measurement tips:
- Annotate deployment dates so before or after comparisons make sense.
- If Assistant or news surfaces show up in referrals, segment them in reporting.
- Be realistic. Valid speakable is not a traffic guarantee; treat it as assistive metadata for discoverability.
Practical guidance for speakable schema
Technical tips:
- Prefer JSON-LD for ease of deployment and maintenance. See JSON-LD if you need a primer.
- Favor simple, resilient selectors (
#story-headline
,.story-summary
) over brittle, deeply nested paths. - Keep each selection to 20 to 30 seconds of speech. Short, clear, and skimmable wins when spoken.
- Do not tag nav, footers, cookie banners, paywall blurbs, or legal text.
- Make sure robots and paywall settings allow the summary to render for crawlers.
- Use either
cssSelector
orxpath
per page, not both.

Content tips:
- Aim for neutral, concise phrasing - think radio news, not product copy.
- Expand acronyms on first mention: “Service level agreement, or SLA...”
- Format numbers and dates for the ear: “twenty-three percent” often plays better than “23%.”
- Skip awkward CTAs or salesy endings in the readout. A clean wrap sounds better.
QA habits that prevent fire drills:
- Re-validate after any template or theme change; small shifts can break selectors. Use the Schema Markup Validator and the Rich Results Test.
- Test selectors in DevTools on desktop and mobile views.
- Watch Search Console for structured data issues and fix them quickly.
- Keep staging and production markup aligned to avoid drift.
- Maintain a lightweight log of pages using speakable and the selectors applied.
- Track the spec’s evolution and check for open issues.
A quick reality check: voice-assistant optimization via speakable can modestly widen your news distribution, but it will not carry a quarterly target on its back. Treat it as an incremental edge for timely stories - product launches, earnings, big conference weeks - when every additional surface helps.
troubleshooting speakable structured data for voice assistants
If voice commands will not surface your article, common culprits include:
- The site is not treated as a news source, reducing eligibility.
- Locale or language is not supported.
- Selectors do not match any visible DOM text at render time.
- The excerpt is too long or awkward when spoken.
- Resources are blocked, or the summary is paywalled without a preview.
- Multiple, conflicting speakable definitions on one page.
A clean debug sequence:
- Validate in the Schema Markup Validator and confirm the speakable property shows the exact selectors you expect. Review Google’s list of structured data errors.
- Use Search Console’s live test to see what Googlebot renders, and check the Unparsable structured data report if parsing fails.
- In DevTools, try:
document.querySelectorAll("#story-headline, .story-summary")
and verify readable text is returned. - Check server logs and core web vitals. If the summary loads late or behind a script gate, crawlers may miss it.
- Confirm canonical tags and that the indexed URL is the one with the markup. If results disappear, see Troubleshoot missing rich results.
- If you suspect policy issues, review the Manual Actions report.
- Stuck? Ask in the Google Search Central forum.
Set expectations accordingly. Even with flawless markup, selection is not guaranteed; ranking and variety matter in Assistant experiences.
If speakable is not a fit for your market or content type, other audio paths can pull weight:
- Launch a podcast or short audio briefings; mark them up with types defined on Schema.org.
- Offer text-to-speech versions of top stories and host the audio with transcripts for accessibility.
- If you qualify as a news source, tune feeds and settings in Publisher Center so your stories are easy to discover.
Final thought for roadmap-heavy B2B teams: speakable is a niche checkbox with a narrow window of upside. If you run a steady news cadence, it is a smart small bet. If not, prioritize SEO work that proves ROI faster, then revisit when your content mix includes timely stories worth being heard.