JSON-LD implementation in Astro, with tested output.

On 6 September 2026, we built a small standalone page with Astro 7.2.2 and Node 26.7.0. It contains the real Marsen name, public website address and a description also shown in the page body. The fixture is an implementation test; it is not a customer deployment or a replacement for the Marsen homepage.

The build emitted one JSON-LD script whose parsed object matched the input. A negative test inserted a second full definition of the same entity and confirmed that the duplicate check rejected it. Another test put an HTML script-closing sequence in the name: Astro still emitted only one script, and parsing the JSON recovered the original text.

• Download the tested Astro page and Organization input.
• Inspect the actual generated HTML.
• Read the timestamped test results and reproduction instructions.

We also checked the existing built Marsen homepage, schema generator and Search product page. Each contained one JSON-LD block and no repeated full entity definitions with the same identifier. This targeted code check does not rule out every possible schema conflict across the website.

Open the current page source and search for application/ld+json. In a CMS, inspect the SEO plugin and theme settings too. In Astro, inspect the shared layout and page component. A layout may already publish the Organization, WebSite and WebPage entities while a guide adds its Article.

Use one maintained definition for an entity on a page. A shared identifier such as https://www.marsentech.com/#organization lets an Article refer to its publisher without repeating the entire organization object. Multiple script blocks are not automatically wrong, and an @id-only reference is not a duplicate definition. The problem is conflicting or needlessly repeated descriptions of the same thing.

Our site already has a layout-owned organization. We tested the example in an isolated fixture rather than adding another Organization script to this guide. Edit the existing owner if one is present. Record who maintains the name, URL, logo and profile links so a rebrand does not leave incompatible versions in a plugin and a template.

Open the schema markup generator, choose Organization and enter the business name and canonical homepage URL. Optional fields should contain real values or remain empty. The generator creates a draft; it does not verify your business, review the page or certify a Google feature.

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://www.marsentech.com/#organization",
  "name": "Marsen Technologies",
  "url": "https://www.marsentech.com/",
  "description": "Marsen connects customer relationships, AI agents and search visibility."
}

The example adds a stable @id to the basic draft for use in other entity references. Replace the Marsen facts with your own when adapting it. Do not invent an address, rating, certification or price to make a validator display another green check.

Organization describes the business identity. An article needs Article properties, while a service page may describe a Service. Use the actual subject instead of attaching every type to every URL. Google’s Organization documentation explains the properties it uses.

In a new minimal Astro project, put the downloaded input in src/pages/organization.json and the page in src/pages/index.astro. The essential implementation is below; the complete download includes visible content and document metadata.

---
import organization from './organization.json';
const jsonLd = JSON.stringify(organization, null, 2)
  .replace(/</g, '\\u003c')
  .replace(/>/g, '\\u003e')
  .replace(/&/g, '\\u0026');
---
<script type="application/ld+json" set:html={jsonLd} />

Astro renders this at build time. It does not need a hydrated React component or a browser event to insert the markup. The Astro directive documentation includes the JSON-LD pattern and explains that set:html does not automatically escape its value. Here serialization is followed by escaping HTML-sensitive characters, preventing a value containing a closing script tag from terminating the element.

Use the same reviewed data for visible content and structured data where practical. Do not concatenate untrusted strings into JSON or paste the object into executable JavaScript. The application/ld+json type identifies a data block. Our fixture is distributed as source and text output, so it does not create a competing public homepage.

Run the production build before judging the result. Inspect dist/index.html, not only the component source. Copy the downloadable duplicate checker to the example project and run:

npm run build
node check-jsonld.mjs dist/index.html

For our fixture, the checker reports one JSON-LD block and an empty duplicateEntityDefinitions list. It parses every matching JSON-LD script and counts full definitions carrying both @id and @type. Repeated references containing only @id are allowed.

This check has limits. Two anonymous objects can describe the same business without sharing an identifier. A plugin can add schema after HTML loads. Microdata can duplicate a JSON-LD entity. Inspect the rendered page and plugin output too; this is a regression check, not a complete semantic validator.

After a template change, repeat the test on a representative homepage, article and product page. Keep the command in the release checklist so a later plugin or layout update cannot quietly reintroduce the same organization.

We pasted the exact generated HTML into the Schema Markup Validator. It detected one Organization with zero errors and zero warnings. This screenshot is the actual code-test result from 6 September 2026.

We tested the same HTML in Google’s Rich Results Test. The initial fixture included noindex, and Google’s code test refused to evaluate it. We saved that failed-test screenshot, removed the directive from the isolated fixture, rebuilt and tested again. The rebuilt example returned “No items detected” in Google’s test, while Schema.org still recognized the Organization. This minimal identity example therefore demonstrates valid vocabulary, not an eligible rich-result feature. Read the recorded validator outcomes.

A passing test does not promise a search appearance. Google’s structured-data guidelines require accurate, relevant, accessible content and leave display decisions to Google. A code test also does not test live fetching or establish that a URL is indexed.

1. Review the facts and choose the existing schema owner.
2. Build and inspect the emitted HTML, including visible content.
3. Run the duplicate check and both appropriate validators.
4. Deploy the intended public page, then validate its live URL.
5. Inspect that URL in Search Console to check access and indexing independently.
6. Save the source version, output, date and screenshots together.

Refresh the evidence when the implementation changes, a relevant validator changes its requirements, or the entity’s public facts change. A new screenshot should represent a new test; changing a date alone does not refresh the implementation.

Measure whether the page gets indexed and reaches the right audience through the AI visibility measurement workflow. Marsen Search connects visibility work to the broader business context. Schema and llms.txt do not provide an automatic ranking or AI-citation benefit.