Fixing Duplicate Schema From the AI Generator

How generated markup ends up doubled

The generator reads one page and writes a complete description of it. It cannot see that your plugin already describes the same page, so it writes its own version of everything it thinks belongs there. Duplicates then arrive by a few routes.

• Generated Article on a plugin-managed post. Yoast SEO, Rank Math and similar plugins output an Article or BlogPosting for posts by default. A generated Article pasted into the post adds a second one.
• Generated Product on a shop page. Many Shopify themes and WooCommerce itself output Product markup on product pages. A generated Product adds a second, and a review app often adds a third with its ratings.
• Site-wide entities repeated per page. The generator may include an Organization, WebSite or BreadcrumbList in its output. Your plugin already writes these on every page.
• The same page generated twice. Someone regenerates markup after an edit and pastes the new block without removing the old one, or pastes it once into the content and once into a header scripts field.
• Tag managers. A block added through Google Tag Manager or another script loader is invisible in the page source, so it is easy to forget and add again elsewhere.

What a clash looks like

A recipe blog on WordPress, fictional Hedgerow Kitchen, has this from its SEO plugin:

{ "@type": "Article", "headline": "Blackberry and apple crumble",
  "author": { "@id": "https://example.co.uk/#/schema/person/1" },
  "datePublished": "2026-08-30T08:00:00+01:00" }

And this from a generated block pasted into the post:

{ "@type": "BlogPosting", "headline": "The best blackberry and apple crumble",
  "author": { "@type": "Person", "name": "Hedgerow Kitchen Team" },
  "datePublished": "2026-08-30" }

Two articles, two headlines, two different authors. Search engines have to decide which to believe, and neither of them is the Recipe markup that would actually help the page.

Why it matters, and what it does not do

Having several JSON-LD blocks on a page is normal and allowed. Google reads them all, and a page can quite properly carry a WebPage, a BreadcrumbList, an Organization and a Recipe from different sources. Google does not penalise duplicate markup as such. The trouble comes when two blocks describe the same thing and disagree.

In short, duplicates waste the effort of generating markup. They make it harder to tell which description is authoritative, and they double the places you have to update when a price, author or date changes.

Finding every block, including hidden ones

View source (Ctrl+U, or Cmd+Option+U on a Mac) and search for application/ld+json to count the blocks in the HTML your server sends. That misses anything added by JavaScript, so also check the rendered page. Open developer tools, go to the Console and run:

document.querySelectorAll('script[type="application/ld+json"]').forEach((s, i) => {
  try {
    const d = JSON.parse(s.textContent);
    const nodes = d['@graph'] || (Array.isArray(d) ? d : [d]);
    console.log(i, nodes.map(n => n['@type']).join(', '));
  } catch (e) { console.log(i, 'unreadable JSON'); }
});

This prints each block's number and the types inside it, whether it came from the server or a script. If the console shows more blocks than view-source did, the extra ones are injected by JavaScript, often from a tag manager or an app.

The Rich Results Test also renders the page and lists detected items by rich result type. Two entries for the same type on a page that should have one is the clearest sign of a clash.

Deciding who owns each type

The durable fix is a simple rule: each type on a page has exactly one source. Decide it once for the whole site, per template, and write it down.

When the generator's version of a type is clearly better than your plugin's, you can reverse the ownership: turn the plugin's output for that type off and keep the generated one. Just do not run both.

The fix, step by step

Merging instead of deleting

Sometimes neither source is complete. The plugin's Article has the right author and dates, and the generated block adds a useful about or mentions list. You can combine them without creating a second article by giving the generated node the same @id as the plugin's node. JSON-LD treats nodes with the same identifier as one entity, so the properties are read together.

A cleaner approach, where the plugin allows it, is to add the extra properties through the plugin's own filter or custom fields so one block contains everything. If you do merge by @id, the two nodes must not disagree: a property given different values in each becomes a conflict, not a merge.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Recipe",
      "@id": "https://example.co.uk/blackberry-apple-crumble/#recipe",
      "name": "Blackberry and apple crumble",
      "mainEntityOfPage": { "@id": "https://example.co.uk/blackberry-apple-crumble/" },
      "author": { "@id": "https://example.co.uk/#/schema/person/1" },
      "publisher": { "@id": "https://example.co.uk/#organization" },
      "image": ["https://example.co.uk/images/crumble-4x3.jpg"],
      "recipeYield": "6 servings",
      "prepTime": "PT20M",
      "cookTime": "PT40M",
      "recipeIngredient": ["500 g cooking apples", "250 g blackberries", "150 g plain flour", "100 g butter", "75 g demerara sugar"],
      "recipeInstructions": [
        { "@type": "HowToStep", "text": "Peel, core and slice the apples, then mix with the blackberries in a baking dish." },
        { "@type": "HowToStep", "text": "Rub the butter into the flour, stir in the sugar and scatter over the fruit." },
        { "@type": "HowToStep", "text": "Bake at 190C for 40 minutes until golden." }
      ]
    }
  ]
}
</script>

This block is what the generator's output looks like after step 4: only the Recipe, with the author, publisher and page linked to the @id values the SEO plugin already publishes. The identifiers shown are placeholders; copy the real ones from your own page source.

Platform notes

WordPress

Yoast SEO, Rank Math and All in One SEO each let you change the schema type per post and switch off article markup for a post type in their settings. Yoast also offers a developer filter, wpseo_json_ld_output, that turns its structured data off entirely, which is only sensible if something else will provide WebSite, WebPage and Organization. Page builders and recipe plugins may add their own blocks too, so check them before assuming the SEO plugin is the only source. Generated blocks pasted in a Custom HTML block are stored in the post content and survive plugin changes, which is why they are easy to forget.

Shopify

Many themes output Product markup in the product template, often in a section or snippet file, and review apps commonly add their own ratings markup. Before adding a generated Product, check the theme code for existing application/ld+json output and the review app's settings for a structured data option. Keep price and availability with whichever source reads live product data, so sales and stock changes are reflected automatically.

Custom and headless sites

Search the codebase for application/ld+json. Layout components often emit site-wide markup that page components then repeat. A single helper that builds one @graph per page, with each type added in one place, prevents duplicates by design.