JSON-LD Patterns That Drive AI Citations: A Technical Playbook

JSON-LD is the language through which your website communicates its structure, authorship, and factual context to AI models. While HTML tells AI crawlers what your content says, JSON-LD tells them what it means — its type, its provenance, its relationships to other content, and the verifiable claims it makes. This distinction is critical for generative engine optimisation because AI models use structured data not merely to index your pages but to evaluate their trustworthiness and citability.

Yet the vast majority of websites dramatically underutilise JSON-LD. Only 12% of UK websites implement more than one schema type per page, according to Aether platform audits conducted in early 2026. This means that 88% of sites are leaving substantial citation potential on the table by failing to provide AI models with the full range of structured signals available to them. This guide provides the specific patterns, combinations, and implementation techniques that drive disproportionate AI citation rates.

JSON-LD Fundamentals for GEO

JSON-LD (JavaScript Object Notation for Linked Data) is a structured data format that embeds machine-readable metadata within a script tag in your page's HTML. Unlike Microdata or RDFa, which interleave structured data within the visible HTML content, JSON-LD sits in a separate block — typically in the document's head section — making it easier to implement, maintain, and debug without affecting the page's visual presentation.

Why JSON-LD Over Other Formats

Google has explicitly recommended JSON-LD as the preferred structured data format, and AI crawlers follow the same preference. JSON-LD's separation from HTML content means it can be managed independently, generated dynamically by server-side code, and automated at scale without touching content templates. For GEO purposes, this separation is particularly valuable because it allows you to iterate on your schema implementation without risking changes to the content that AI models are already citing.

The format also supports nesting and referencing, which enables complex entity relationships that simpler formats cannot express. You can describe a BlogPosting that has an author who is part of an Organisation that offers a SoftwareApplication — all within a single, parseable JSON-LD block. This depth of entity description is precisely what AI models use to build comprehensive understanding of your content's context and authority.

The Dual-Schema Strategy

The single most impactful JSON-LD pattern for AI citations is the dual-schema strategy: implementing two complementary schema types on a single page. Aether Research data from 2026 shows that BlogPosting combined with FAQPage schema increases AI citations by 189% compared to BlogPosting alone. This dramatic improvement occurs because the two schema types serve different functions in the AI evaluation process.

BlogPosting provides the foundational context: what the content is, who wrote it, when it was published, and what topic it covers. FAQPage provides extractable question-answer pairs that AI models can directly incorporate into their responses. Together, they give AI models both the authority signals to trust your content and the structured data to cite it efficiently.

The schema vocabulary was designed to be combinatorial. Using a single type per page is like writing a book with only nouns. The real power emerges when types are layered to express the full dimensionality of what a page represents.

Aaron Bradley — Schema.org expert (paraphrased)

The 6 Schema Types That Drive Most AI Citations

Through analysis of citation patterns across ChatGPT, Perplexity, Google AI Overviews, and Claude, we have identified six JSON-LD schema types that account for the vast majority of AI citations. Each serves a distinct purpose, and the most effective implementations combine two or more of these types on a single page.

1. BlogPosting

BlogPosting is the foundation schema for any content-driven page. It communicates the article's headline, author, publisher, publication date, word count, and primary topic. For AI models, BlogPosting schema serves as a credentials check — it verifies that the content is a published, authored work from a named organisation, not an anonymous or auto-generated page. Every article, guide, and insight piece on your site should include BlogPosting schema with all required and recommended properties populated.

2. FAQPage

FAQPage schema is the highest-impact type for direct AI citation. It presents questions and answers in a structured format that AI models can extract verbatim. When an AI model encounters a query that aligns with one of your FAQ questions, the structured answer provides a ready-made response unit. This is why the BlogPosting plus FAQPage combination is so powerful — the BlogPosting provides authority context, and the FAQPage provides extractable content units.

3. HowTo

HowTo schema structures step-by-step instructions in a format optimised for procedural queries. According to Semrush (2026), content with HowTo schema earns 2.4 times more step-by-step AI answer inclusions than unstructured instructional content. If your content explains how to do something — implement a strategy, configure a tool, execute a process — HowTo schema should accompany your BlogPosting schema to maximise citation potential for procedural queries.

4. Product

Product schema communicates structured information about software tools, services, or physical products. For SaaS businesses, Product schema enables AI models to accurately describe your offering's features, pricing, and competitive positioning when responding to comparison or recommendation queries. This is particularly relevant for businesses that want AI models to include their product in response to queries like "best tools for" or "alternatives to" specific solutions.

5. LocalBusiness

LocalBusiness schema is essential for any business with a physical presence or service area. AI models increasingly field local queries — "best accountant in Surrey," "web agency near me" — and LocalBusiness schema provides the structured data needed to include your business in these responses. Combined with answer engine optimisation techniques, LocalBusiness schema positions your brand for the rapidly growing category of AI-powered local search.

6. Organization

Organization schema establishes your business entity in the knowledge graph that AI models build internally. It communicates your company name, URL, logo, social profiles, and founding details. While it does not directly trigger citations for individual content pieces, it strengthens the overall authority signal that AI models associate with your domain, making all of your content more likely to be cited. Think of Organization schema as the foundation upon which content-level citations are built.

Advanced Patterns and Nesting Techniques

Beyond basic implementation, advanced nesting and relationship patterns can significantly amplify the citation impact of your structured data. These techniques are used by fewer than 5% of websites, representing a substantial competitive advantage for early adopters.

Entity Nesting for Authority Chains

Rather than using flat, standalone schema blocks, nest related entities within your primary schema to create authority chains. For example, within your BlogPosting schema, nest the author as a Person entity with their own credentials (jobTitle, worksFor, sameAs links to professional profiles). Within the publisher, nest the Organisation entity with its own URL, logo, and founding details. Each level of nesting adds a new layer of verifiable context that AI models can use to assess your content's trustworthiness.

The nesting pattern extends to content relationships. If your article references other articles on your site, you can use the citation property within BlogPosting to formally declare those relationships. This creates a structured content graph that mirrors the internal linking on your page, reinforcing the semantic connections through both HTML links and schema declarations.

Combining HowTo with BlogPosting and FAQPage

The triple-schema combination of BlogPosting, FAQPage, and HowTo on a single page is the most powerful pattern we have identified for instructional content. The BlogPosting provides the authority frame. The FAQPage addresses the conceptual questions that readers ask before attempting the process. The HowTo structures the actual steps in a format optimised for procedural extraction. According to our analysis of Google AI Overview citation patterns, this triple combination achieves citation rates approximately 2.7 times higher than single-schema implementations for instructional content.

The @graph Pattern for Multi-Entity Pages

When implementing multiple schema types on a single page, the @graph pattern provides a clean, specification-compliant way to declare all entities within a single JSON-LD block. Rather than using multiple separate script tags, you wrap all your entities in a @graph array within one block. This approach is technically cleaner, reduces the chance of conflicting declarations, and makes it easier for validation tools to assess the complete schema output of a page.

The businesses achieving the highest AI citation rates are not those with the most content or the highest domain authority. They are the ones providing AI models with the richest, most accurate structured data. JSON-LD is the competitive moat that most businesses do not even realise exists.

Aether Insights

Implementation Examples for Common CMS Platforms

Implementing advanced JSON-LD patterns varies by CMS platform, but the core principles remain constant. Below, we outline the recommended approach for the most common platforms used by UK businesses.

WordPress

WordPress offers the broadest ecosystem of schema plugins, but plugin-generated schema rarely includes the dual-schema patterns that maximise AI citations. The recommended approach is to use a reliable plugin like Yoast or Rank Math for base BlogPosting schema, then supplement with custom template code in your theme's header.php or a custom plugin for FAQPage and HowTo blocks. This hybrid approach gives you the convenience of plugin-managed base schema with full control over advanced patterns.

For sites with hundreds of articles, consider building a custom shortcode or Gutenberg block that generates FAQPage JSON-LD from structured content fields. This allows content editors to add FAQ schema through the familiar WordPress editing interface without needing to write JSON-LD manually. Ensure that your structured data testing workflow validates the output of both the plugin and the custom implementations to catch any conflicts.

Shopify and E-Commerce Platforms

Shopify's built-in schema is limited to basic Product schema. For GEO optimisation, supplement it with BlogPosting schema on your blog content, FAQPage schema on product pages that include FAQ sections, and Organization schema site-wide. Shopify's Liquid templating system allows you to inject JSON-LD blocks into theme templates, though the syntax requires more careful handling than PHP-based systems.

Headless CMS and Static Sites

Headless CMS platforms (Contentful, Sanity, Strapi) and static site generators (Next.js, Gatsby, Hugo) offer the most flexibility for JSON-LD implementation because schema can be generated programmatically from content models. Define your schema patterns once in code, map them to your content fields, and every page automatically receives correct, comprehensive structured data. This is the gold standard approach and the one we recommend for any new build.

---

The structured data landscape for AI visibility is evolving rapidly. Schema types that had minimal impact in traditional SEO are now driving measurable citation gains in AI search. The businesses that invest in comprehensive, multi-layered JSON-LD implementations today are building a competitive advantage that will compound as AI models become increasingly sophisticated in their use of structured signals.

Start with the dual-schema foundation of BlogPosting plus FAQPage, then expand to include HowTo, Product, and Organization schemas where appropriate. Test rigorously, monitor continuously, and iterate based on the citation data you observe. The technical investment is modest, but the visibility returns are substantial and growing.