Skip to content
CiteScore

Speakable schema

Free Speakable schema generator

21 October 2026 · CiteScore Team · 8 min read

Free AI Answer Snapshot

When customers ask ChatGPT or Perplexity for your trade in your city, who do they name?

Get a free AI Answer Snapshot
Get the Fix Pack · $49 (opens in a new tab)

We email the Fix Pack within a few business days after payment, once we have the business name, city, service, and public site URL. The re-check in the pack is the same prompts, run again 14 days later. Full refund within 14 days of delivery if the Fix Pack is not useful. Refund policy.

Opens Payhip in a new tab. The shop name on that page is DeskOps Templates (opens in a new tab). Payhip lists a text file with the order. We email the Fix Pack. What happens after you buy.

The free AI Answer Snapshot asks logged-out ChatGPT and Perplexity and records the answer. The Fix Pack is $49. This generator copies facts the page already shows. It does not make an assistant name the business.

AI answers vary run to run. This is a snapshot, not a ranking, and not a promise that any assistant will name the business.

See an example snapshot · See what's in the Fix Pack. Example pages use illustrative numbers. Not a customer result.

Prepared by the CiteScore Team.

Request a free AI Answer Snapshot

Same request as the free snapshot page. AI answers vary run to run. This is a snapshot, not a ranking, and not a promise that any assistant will name the business.

All fields are required.

We email the snapshot to this address. We do not ask for passwords, cookies, or content behind a login. We do not sell emails. Privacy.

Email this request instead
citescore@mail.grokbot.com

No mail app opened? Copy this and email it to citescore@mail.grokbot.com from any inbox.

If that send fails, email this request instead, or use the copy buttons.

Sends the business name, city, service, website, and email to citescore@mail.grokbot.com through FlowForm. This page path is included. No account and no paid checkout.

Prepared by the CiteScore Team.

Free, no-signup SpeakableSpecification JSON-LD generator. Add a page name and the CSS selectors or XPath for a short passage already on the page. Copy the WebPage script. Nothing is uploaded.

The script updates as you type. Name the page, then point at a short passage a person can already read. Nothing is uploaded.

CSS selectors

One locator per row. A blank row is left out. Duplicates are written once. Google’s example uses about two or three sentences, or around 20–30 seconds of speech, and it says not to mark the whole article.

Speakable JSON-LD

Generated in your browser. Nothing is uploaded.

Get a free AI Answer Snapshot · Fix Pack, $49 · Run the free check

When a passage is speakable

A news page has a headline and a short summary a person can read. A text-to-speech system still has to decide which block to speak and which block to skip. schema.org speakable is that decision written down. The value is a SpeakableSpecification with cssSelector or xpath, or a URL that points at an element id. A parser can read the locator without guessing which paragraph was meant to be heard. The script does not create a citation, and it does not rank the page. A law firm’s office page is a different problem, on AI search visibility for law firms. A clinic or health-tech site is on AI search visibility for healthcare. A brokerage office is on AI search visibility for real estate. A restaurant location is on AI search visibility for restaurants. An accounting office is on AI search visibility for accountants. A dental office is on AI search visibility for dentists. A plumbing business’s office, service area, and hours are on AI search visibility for plumbers. An HVAC company’s office, service area, and hours are on AI search visibility for HVAC. A roofing contractor’s office, service area, and hours are on AI search visibility for roofers. An electrical contractor’s shop, service area, and hours are on AI search visibility for electricians. A landscaping business’s office, service area, and hours are on AI search visibility for landscapers. An auto repair business’s shop, service area, and hours are on AI search visibility for auto repair. A chiropractic office’s office, services, and hours are on AI search visibility for chiropractors. An insurance agency’s office, lines of coverage, and hours are on AI search visibility for insurance agents. A veterinary clinic’s office, services, and hours are on AI search visibility for veterinarians. A photo credit uses the ImageObject schema generator.

Google’s speakable documentation, read on 21 October 2026, says the feature is in beta. The Google Assistant uses the markup to answer topical news queries on smart speaker devices. It returns up to three articles from around the web, reads the marked sections with text-to-speech, attributes the source, and sends the full article URL to the user’s mobile device through the Google Assistant app. The same page limits that behavior to users in the U.S. who have Google Home devices set to English, and to publishers that publish content in English. It says the property works with Article or Webpage. schema.org says Article or WebPage. This form writes WebPage. An article that also needs a headline, a date, and a byline uses the Article schema generator. Add speakable to that script by hand if the page is an article and the passage is visible. This form does not write those article fields.

The pages that document the property do not name ChatGPT, Perplexity, Claude, Gemini, Copilot, or Google AI Overviews as consumers of speakable. schema.org describes the property for text-to-speech in general. Google’s page describes one beta, for news, on Google Assistant. A selector on a product page or a firm homepage does not turn that page into the news case. Google’s troubleshooting note says that if a voice command does not play the story, ranking is algorithmic, and the Assistant provides up to three articles from different publications.

What the script includes

The script is written once the page name and at least one valid locator are present. name is the name a person can see. url is included when it is an http or https URL. A missing scheme is read as https. An invalid URL is left out, and the name can still be written.

Google’s table says to use either cssSelector or xPath, and not both. schema.org’s property is xpath. The letters differ. This form writes one SpeakableSpecification. The locator you do not select is left out, even if those rows still have text. When the locator is XPath, the property name is the spelling you pick: xpath for schema.org, or xPath for the key on Google’s page. The form does not write both spellings. One locator is a one-item array. More than one is a longer array. That is the shape of the JSON-LD examples on both pages. A repeated row is written once. A blank row is left out.

schema.org’s JSON-LD example uses cssSelector values headline and summary, without a dot. Google’s example uses .headline and .summary. This form keeps the text you type, so a class selector keeps its dot and a bare class name stays bare. Match the class or the id that is actually on the page. Google’s XPath example is /html/head/title and /html/head/meta[@name='description']/@content. An expression has to start with /, ./, ../, or (. A class selector typed into the XPath rows is left out.

Google’s content guidelines say the marked text should be a concise headline or summary, and they recommend around 20–30 seconds per section, or roughly two or three sentences. The technical guidelines say not to mark a dateline, a photo caption, or a source attribution, because those sound confusing when only the voice is heard, and not to highlight the entire article. This form leaves out the CSS selectors html, body, :root, and *, and the XPath values /html and /html/body. It cannot see your template, so a selector such as article is kept when you type it. Use it only when that element is the short passage, not the whole story.

schema.org also allows speakable to be an id-value URL, such as #summary, with no SpeakableSpecification. This form does not write that shape. Use a CSS id selector, #summary, when the element has that id and you want the specification type. Google’s structured data policies, read on 21 October 2026, say not to mark up content that is not visible to readers of the page, and that the structured data must be a true representation of the page content. The copy block is one script tag. Paste it into the HTML the server sends for that page. Answer engines often skip text that only appears after JavaScript runs. The free check lists the JSON-LD @type values on the homepage. A news URL at another path is outside that fetch. The GEO audit looks for markup that matches the copy. The longer note on which facts to repeat is the schema guide.

How to install it

  1. Copy the script.
  2. Paste it into the HTML the server sends for that page, in the head or the body.
  3. Confirm the same name is visible, and that each locator matches text a person can already read.
  4. If the page is English news you want the Google Assistant beta to read, choose the xPath spelling before you copy. schema.org’s spelling is xpath.

What to leave out

Questions about Speakable schema

Does Speakable schema guarantee a Google Assistant readout or an AI citation?

No. Google’s speakable documentation, read on 21 October 2026, says the feature is in beta and subject to change, and that Google does not guarantee that features which use structured data will show. The documented consumer is the Google Assistant answering topical news queries on Google Home devices, for users in the U.S. with the device set to English, from publishers that publish in English. The Assistant returns up to three articles and attributes the source. The script does not make ChatGPT, Perplexity, Claude, Gemini, Copilot, Grok, Meta AI, DeepSeek, Mistral, Brave Search, You.com, DuckDuckGo, or Google AI Overviews cite the page. Those products are not named on schema.org’s speakable page or on Google’s speakable page.

When does speakable markup help a voice assistant?

schema.org’s speakable property, read on 21 October 2026, marks sections of a Web page that are especially appropriate for text-to-speech. It is used on Article or WebPage. Google’s narrower beta uses that markup so the Google Assistant can read a short news passage aloud. The helpful case is an English news page whose headline or summary is already visible, concise, and not a dateline, a photo caption, or a source attribution. Google recommends around 20–30 seconds per section, or roughly two or three sentences, and says not to highlight the entire article. A product page, a local-business page, or a company FAQ is outside the news case that page describes.

Should the locator be cssSelector or xpath, and which spelling?

Google’s speakable page says to use either cssSelector or xPath, and not both. schema.org’s property name is xpath. Google’s page spells that key xPath. This form writes one SpeakableSpecification and one locator. When you choose XPath, you also choose the spelling. schema.org also allows a speakable value that is an id-value URL, such as a fragment. This form does not write that shape. A single locator is still written as an array, which is the shape of the examples on both pages.

Where should the script go?

On the page whose passage you are marking, in the HTML the server sends, where a person can already see that text. Google’s structured data policies say not to mark up content that is not visible to readers of the page. The script is built in the browser. Nothing is sent to CiteScore.

Get a free AI Answer Snapshot · Fix Pack, $49 · Run the free check

A separate GEO audit PDF covers one public URL.