Schema Markup in the Age of AI
Schema markup used to be about earning rich snippets in Google — star ratings, FAQ toggles, recipe cards. Google has since removed many of those SERP features, and a common misconception is that schema no longer matters. The opposite is true. Schema’s importance has shifted and intensified because large language models can parse structured data with far greater accuracy than unstructured prose.
When ChatGPT, Perplexity or Google AI Overviews generate an answer about your business, they need to determine facts: who you are, what you do, where you operate, what services you offer, what your clients say. If those facts exist as JSON-LD structured data on your site — with explicit types, properties and relationships — the AI extracts them with high confidence. If those facts are buried in marketing copy, the AI has to infer them, and inference introduces error, ambiguity and the risk of being passed over entirely.
Schema markup is the bridge between traditional search optimisation and LLM Optimisation. It serves both Google’s crawlers and AI training pipelines simultaneously. For businesses serious about visibility in 2026 and beyond, comprehensive structured data isn’t optional — it’s foundational infrastructure.
What Is Schema Markup (JSON-LD)?
Schema markup is a standardised vocabulary (defined at Schema.org) that describes the entities, relationships and facts on a web page in a format machines can reliably parse. JSON-LD (JavaScript Object Notation for Linked Data) is the implementation format Google recommends — it sits in a <script> tag in your page’s HTML, separate from visible content, and declares structured facts about the page and the entities it describes.
A simple example: your homepage might contain JSON-LD that declares “this website belongs to a ProfessionalService entity called SEO Strategy Ltd, founded in 2005, based in Southampton, Hampshire, that knows about Search Engine Optimisation and LLM Optimisation.” That single block gives search engines and AI systems a definitive, machine-readable fact sheet — no scraping, no inference, no ambiguity.
Schema works through a graph model. Individual nodes (Organization, WebPage, FAQPage, Person, Service) connect via typed relationships (publisher, mainEntityOfPage, provider, author). When implemented correctly, your site’s structured data forms a coherent knowledge graph that mirrors how search engines and LLMs organise information internally. This is why schema and entity SEO are inseparable — schema is how you declare your entity to machines.
Schema Types That Matter for SEO and AI Visibility
Not all schema types carry equal weight. Some directly trigger Google rich results. Others are invisible in SERPs but highly valuable for AI systems. The most effective schema implementations prioritise types that serve both audiences. These are the types we implement for clients, categorised by their primary value:
Entity Foundation Types
Organization / ProfessionalService / LocalBusiness — Establishes your core entity identity: name, address, founding date, founder, contact details, social profiles (sameAs), areas served, and topics of expertise (knowsAbout). This is the single most important schema type for both Google Knowledge Panel eligibility and AI entity recognition. Without it, search engines and LLMs have to guess who you are from scattered signals. With it, you’re declaring it definitively.
Person — Establishes individual entity authority for founders, authors and key team members. Crucial for E-E-A-T signals and AI author attribution. When your blog posts carry Author schema linked to a Person entity with credentials, affiliations and sameAs profiles, both Google and LLMs can evaluate the expertise behind your content.
WebSite / WebPage — Structural schema that tells machines what type of content each page contains and how pages relate to each other and to your Organization entity. The @id graph connections between WebSite → WebPage → Organization create the backbone of your structured data architecture.
Content Enrichment Types
FAQPage — Explicit question-answer pairs that AI systems can extract and cite with maximum confidence. Even though Google removed FAQ rich results for most sites, FAQPage schema remains one of the most valuable types for LLM visibility. When an AI encounters a well-formed FAQ with a clear question and authoritative answer, it has a ready-made citation. We consider FAQPage the single highest-value schema type for Answer Engine Optimisation.
HowTo — Step-by-step processes with named steps, descriptions and optional images. Valuable for instructional queries where AI systems need to present sequential information. HowTo schema makes your processes machine-parseable rather than requiring AI to extract steps from prose — reducing inference errors and increasing citation likelihood.
Article / BlogPosting — Content metadata including headline, author, publisher, dates, description and word count. Establishes provenance (who wrote this, when, for which publication) that both Google and LLMs use for trust evaluation. The combination of Article schema with linked Person and Organization entities creates a complete attribution chain.
VideoObject — Video metadata including title, description, thumbnail, upload date, duration and embed URL. Enables video rich results in Google and provides AI systems with structured information about video content they can’t otherwise parse. Essential if video is part of your content strategy.
Commercial and Specialist Types
Service — Declares what services your business offers, linked to your Organization entity via the provider property. Valuable for AI systems evaluating whether your business is relevant to a user’s commercial query. Service schema with detailed descriptions, area served and audience helps LLMs match your business to “find me a [service] in [location]” style queries from AI agents.
Product / SoftwareApplication — For e-commerce and SaaS businesses: name, price, availability, reviews, features and specifications. Google still supports rich results for Product schema (price, availability, ratings). For AI, Product schema provides the structured facts needed to include your products in comparison responses and recommendations.
Review / AggregateRating — Social proof in machine-readable format. Star ratings and review counts still trigger Google rich results and provide AI systems with quantified trust signals. When an LLM is comparing options for a user, explicit rating data carries more weight than unstructured testimonial text.
BreadcrumbList — Navigation hierarchy that helps search engines understand your site’s information architecture. Also aids AI systems in understanding content relationships — a page nested under /llm-optimisation/geo/ signals its topical context more clearly than a flat URL structure alone.
Event / Course — For businesses running events, webinars, training or courses. Event schema triggers rich results with dates, locations and registration links. For AI, it provides structured temporal data that enables “what events are coming up about [topic]” style answers.
Why Most Schema Implementations Fail
After twenty years of implementing structured data, the problems we see most often aren’t missing schema — they’re broken schema. Common failures include duplicate Organization nodes (SEO plugin and theme both outputting conflicting versions), empty FAQPage schema (the markup exists but questions and answers are blank), missing @id references (schema nodes that float disconnected from the page graph), incomplete required properties (VideoObject without thumbnailUrl, Product without price), and malformed JSON-LD that silently fails validation.
The most damaging pattern is duplication. If your site outputs two Organization schema blocks with different addresses, phone numbers or names, you’re actively confusing the systems you’re trying to inform. We audit every client site for schema conflicts between theme code, SEO plugins (RankMath, Yoast, AIOSEO), page builders and custom code before implementing any new markup.
Our approach is programmatic: structured data compiles from defined fields, validates automatically, de-duplicates with plugin output via filters, and uses hardened JSON encoding to prevent injection vulnerabilities. The result is clean, valid, non-duplicated schema that passes Google’s Rich Results Test and gives AI systems exactly the structured facts they need.
Schema as the Bridge Between SEO and AI Visibility
This is the insight most consultancies miss: schema markup isn’t just an SEO tactic — it’s the connective layer between your website and every machine that reads it. Google’s crawler, Bing’s indexer, ChatGPT’s browsing tool, Perplexity’s retrieval system, and autonomous AI agents all parse JSON-LD. It’s the closest thing to a universal API for your business information.
When we implement schema for clients, we’re not just thinking about today’s Google. We’re building the machine-readable foundation that serves Generative Engine Optimisation, Answer Engine Optimisation, AI Overview Optimisation, and AI Agent Optimisation simultaneously. A ProfessionalService entity with comprehensive knowsAbout properties serves Google’s Knowledge Graph today and an AI agent’s vendor evaluation tomorrow.
This is also why schema connects directly to entity SEO and technical SEO. Entity SEO defines what your business is and what it’s authoritative for. Schema declares those facts in machine-readable format. Technical SEO ensures the infrastructure that delivers those facts is crawlable, fast, and correctly configured. Together, these three disciplines create what we call a Governed Knowledge System — your business’s machine-readable truth layer.
Our Schema Implementation Process
For a hands-on, code-first walkthrough of JSON-LD syntax, worked examples for every major schema type, the @id graph model, WordPress implementation patterns, and testing tools, see our companion JSON-LD Implementation Guide.
We don’t bolt schema onto existing sites as an afterthought. We architect it as core infrastructure. Every implementation follows the same systematic process: audit existing structured data and identify conflicts, define the entity graph (what entities exist, how they relate), implement JSON-LD with stable @id references and proper graph connections, de-duplicate with SEO plugin output, validate across Google Rich Results Test and Schema.org validator, and establish ongoing monitoring.
For WordPress sites, we implement schema at the theme level — compiled from structured meta fields, not generated by plugins that can conflict or be overridden. This gives us complete control over output quality, eliminates duplication, and ensures schema survives plugin updates and configuration changes. The result is structured data that works as hard as your content does — serving search engines, AI systems and future platforms you haven’t encountered yet.
If you want to understand where your current schema stands, our AI Visibility Audit includes a comprehensive structured data analysis alongside AI citation testing and entity evaluation.