WordPress Structured Data (JSON-LD)

Model note: Building a single schema type is mechanical (haiku). Reach for sonnet/opus only when reconciling a complex graph against an existing SEO-plugin graph.

Emit schema.org JSON-LD from theme/plugin code for content an SEO plugin can't generate on its own — without duplicating what the SEO plugin already ships.

When to use

• Content lives in post meta / custom fields (e.g. an FAQ repeater) and is stripped from the rendered content, so Rank Math/Yoast can't detect it.
• You render a custom section (table of contents, related items, steps) that warrants ItemList / HowTo.
• You need a schema type the SEO plugin doesn't offer on that template.

Rule 0 — never duplicate the SEO plugin's graph

Most sites already run Rank Math, Yoast, or SEOPress. They emit a @graph with WebSite, WebPage, Organization, Person, BreadcrumbList, and an article type (Article/BlogPosting/Product). Never re-emit those — duplicate/competing nodes confuse parsers and can suppress rich results.

Always check first what is already on the page:

# View source, then list the @type values in every ld+json block
curl -s "<url>" | grep -A99 'application/ld+json'

Or in the browser console:

[...document.querySelectorAll('script[type="application/ld+json"]')]
  .flatMap(s => { const j = JSON.parse(s.textContent); return (j['@graph']||[j]).map(n => n['@type']); });

Only add types the existing graph is missing (commonly FAQPage, HowTo, Recipe, custom ItemList).

Pattern — build in PHP, print on wp_head

add_action( 'wp_head', function () {
    if ( ! is_singular( 'post' ) ) {
        return;
    }

    $faqs = get_post_meta( get_the_ID(), 'my_faqs', true ); // stripped from content
    if ( empty( $faqs ) || ! is_array( $faqs ) ) {
        return;
    }

    $questions = array();
    foreach ( $faqs as $faq ) {
        if ( empty( $faq['question'] ) ) {
            continue;
        }
        $questions[] = array(
            '@type'          => 'Question',
            'name'           => wp_strip_all_tags( $faq['question'] ),
            'acceptedAnswer' => array(
                '@type' => 'Answer',
                'text'  => wp_kses_post( wpautop( $faq['answer'] ?? '' ) ),
            ),
        );
    }

    if ( ! $questions ) {
        return;
    }

    $schema = array(
        '@context'   => 'https://schema.org',
        '@type'      => 'FAQPage',
        'mainEntity' => $questions,
    );

    echo '<script type="application/ld+json">' . wp_json_encode( $schema ) . '</script>' . "\n";
} );

Key points

• Build a PHP array, then `wp_json_encode()` — never hand-concatenate JSON. wp_json_encode() escapes for the <script> context; output is safe to print as-is. (PHPCS's EscapeOutput sniff doesn't recognise it as an escaper; suppress it via a centralized phpcs.xml.dist exclude scoped to the file rather than scattering inline ignores.)
• `@type` text: wp_strip_all_tags() for short names/titles; wp_kses_post() for answer/description bodies that may carry HTML.
• head vs footer is irrelevant to Google — it parses JSON-LD anywhere. Use wp_head to sit alongside the SEO plugin's graph (convention), or wp_footer if you need late data; both work.
• Gate to the right context (is_singular(), post type, a feature toggle) so schema only appears where the content does.
• Mirror visible content. Schema must reflect what users actually see on the page (Google's structured-data policy). Don't emit FAQ schema for FAQs you don't render.

Common types worth emitting from code

Validate

• Google Rich Results Test (search.google.com/test/rich-results) — confirms eligibility per type.
• Schema.org validator (validator.schema.org) — structural correctness.
• Confirm there's exactly one node per type across all blocks (no duplicate Article/FAQPage).

Common Mistakes