Fixing Schema for AI Answers

What schema can and cannot do for AI answers

It helps to be clear about the evidence. Google says it uses structured data to understand page content and the entities on it, and that AI Overviews are built on its normal Search systems. Microsoft recommends structured data in its Bing webmaster guidance. Beyond that, AI companies have said very little about whether their assistants read JSON-LD directly when writing an answer. Any claim that a particular schema property "gets you cited" goes further than the public evidence.

What schema does reliably is remove ambiguity. It can state, in a form software does not have to guess at:

• the exact name of the business and its official website;
• which social and directory profiles belong to it;
• who wrote a page and what that person's role is;
• when a page was published and last changed;
• how the page, the business and the author relate to one another.

Those are the facts that go wrong in AI answers when they are left for a model to infer. Schema does not replace a clear answer in the visible text. Assistants quote what readers see, so the page still has to answer the question well.

One connected graph instead of loose blocks

Many sites have several separate JSON-LD blocks that never refer to each other: a theme outputs an Organization, a plugin outputs an Article with the author as a plain name, and a reviews widget outputs a LocalBusiness with a slightly different name. To a machine these look like three unrelated things.

The fix is to give each entity an @id, a URL-style identifier that stays the same on every page, and to point to it wherever the entity is mentioned. The Article says its publisher is https://www.example.co.uk/#organization; every page that mentions the business uses that same identifier. Now there is one business, described once, referred to many times.

A working example

This graph is for a blog post on a fictional firm, Brookfield Garden Rooms. It is valid JSON and can be adapted.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Organization",
      "@id": "https://www.example.co.uk/#organization",
      "name": "Brookfield Garden Rooms Ltd",
      "alternateName": "Brookfield Garden Rooms",
      "url": "https://www.example.co.uk/",
      "logo": "https://www.example.co.uk/images/logo.png",
      "description": "Designs and builds insulated garden rooms and garden offices in West Yorkshire.",
      "address": {
        "@type": "PostalAddress",
        "streetAddress": "Unit 4, Mill Lane",
        "addressLocality": "Wakefield",
        "postalCode": "WF1 1AA",
        "addressCountry": "GB"
      },
      "sameAs": [
        "https://www.facebook.com/example-garden-rooms",
        "https://www.linkedin.com/company/example-garden-rooms"
      ]
    },
    {
      "@type": "Person",
      "@id": "https://www.example.co.uk/about/sam-carter/#person",
      "name": "Sam Carter",
      "jobTitle": "Lead designer",
      "url": "https://www.example.co.uk/about/sam-carter/",
      "worksFor": { "@id": "https://www.example.co.uk/#organization" }
    },
    {
      "@type": "WebPage",
      "@id": "https://www.example.co.uk/blog/garden-room-planning/#webpage",
      "url": "https://www.example.co.uk/blog/garden-room-planning/",
      "name": "Do garden rooms need planning permission?",
      "isPartOf": { "@id": "https://www.example.co.uk/#website" }
    },
    {
      "@type": "WebSite",
      "@id": "https://www.example.co.uk/#website",
      "url": "https://www.example.co.uk/",
      "name": "Brookfield Garden Rooms",
      "publisher": { "@id": "https://www.example.co.uk/#organization" }
    },
    {
      "@type": "Article",
      "@id": "https://www.example.co.uk/blog/garden-room-planning/#article",
      "headline": "Do garden rooms need planning permission?",
      "mainEntityOfPage": { "@id": "https://www.example.co.uk/blog/garden-room-planning/#webpage" },
      "author": { "@id": "https://www.example.co.uk/about/sam-carter/#person" },
      "publisher": { "@id": "https://www.example.co.uk/#organization" },
      "datePublished": "2026-03-02T09:00:00+00:00",
      "dateModified": "2026-08-14T10:30:00+01:00"
    }
  ]
}
</script>

The name, address and people here are invented. Use your own real details, and only list profiles in sameAs that you actually control.

The properties that carry identity

Properties such as knowsAbout are valid schema.org and can describe specialisms, but there is no public evidence that assistants give them special weight. Add them if they are accurate; do not expect them to change results on their own.

The fix, step by step

Finding what your site outputs now

Most sites already have some markup, and most of the work is tidying it rather than starting again. Check a small sample: the home page, the about or contact page, one service or product page and one blog post.

• Count the blocks. In the page source, count application/ld+json scripts. More than one is fine if they refer to each other by @id; separate unconnected blocks are the problem.
• Look for plain-text authors. "author": "Sam" or "author": {"@type": "Person", "name": "Admin"} tells a machine almost nothing. "Admin" is a common default in WordPress.
• Compare names. Does the Organization name in markup match the name on your Business Profile and invoices? Trading names and registered names often differ; use legalName for the registered one.
• Check dates. Some themes set dateModified to the current time on every page load, which makes every page look freshly updated and the date meaningless.

Platform notes

WordPress

Yoast SEO and Rank Math both output a connected graph with @id values already. The common gaps are settings, not code: the site representation set to a Person when it should be an Organization, no logo, empty social profile fields, and authors left as "admin". Fill in the Organization details in the plugin's site settings and each author's user profile. Avoid adding a second schema plugin, which usually creates duplicate nodes.

Shopify

Themes output their own Organization and Product markup, and many apps add more. Check the theme's theme.liquid and product templates for JSON-LD, and each installed SEO or review app's settings, then choose one source for the Organization. If you add a blog author graph by hand, reference the same Organization @id the theme uses rather than creating a new one.

Custom and static sites

Generate the graph in one shared template or component, fed by one configuration file holding the business details and one data file for authors. Build the JSON with a proper encoder, such as json_encode() in PHP or JSON.stringify() in JavaScript, so a quote mark in a name can never break the block.

Keeping it accurate

• One source of business details. Store the name, address, phone and profiles once and output them everywhere from there, so a change of address updates the site and the markup together.
• Review after plugin and theme updates. Updates are the usual cause of a new duplicate Organization or a reset author setting.
• Add new authors properly. Create the profile page and Person node when someone starts writing, not months later.
• Remove leavers carefully. Keep the profile of a former author if their articles remain, updated to say they have left, so the articles still point to a real page.

For assistants stating wrong facts about your business, see fixing wrong attribution. For the wider schema checks, see the AI Schema guide.