Schema Markup for AEO: What Actually Moves the Needle

6 September 2026 · AxionIQ · seo / ai-search / structured-data / content-marketing

Most schema markup advice for AEO is written by people applying 2019-era SEO thinking to a 2026 problem. The advice to “just add FAQPage schema to everything” or “use every schema type you can justify” produces websites covered in structured data decoration that AI systems ignore entirely. The reason is simple: AI search systems do not use schema markup the way Google does. The rules are different, the signals are different, and the types that produce citations in ChatGPT or Perplexity are not the types that move organic traffic in Google.

This piece is a working breakdown of schema markup for AEO from the perspective of what demonstrably changes citation behaviour in AI search systems. It covers the four schema types that have consistent, observable impact, the six types that are widely recommended and almost entirely irrelevant for AEO, and the implementation approach that produces clean markup without the bloat.

The wrong approaches to schema markup for AEO

The most common wrong approach is treating schema markup as a checklist. Add Organisation, add WebSite, add Article, add FAQPage, maybe throw in some BreadcrumbList, and call it done. This is the approach that produces markup files with twelve node types, zero of which are actually cited by AI systems because the content they annotate is generic boilerplate that AI systems have already discounted.

The second wrong approach is FAQPage on every page. FAQPage schema was designed to produce rich snippets in traditional search. It does produce them in Google. It does not reliably produce citations in AI search systems when the FAQ content is thin, generic, or written specifically to game the markup rather than to answer a real question. AI systems are improving at detecting and discounting FAQ content that exists primarily to capture markup rather than to serve the reader, and the sites that have been most aggressive with FAQ schema are now the ones most likely to be ignored when the content does not match the markup depth.

The third wrong approach is chasing SpeakableSpecification. The theory is that marking content as speakable for voice assistants will produce AEO citations. In practice, speakable markup has minimal observable impact on AI citation patterns, because AI systems select content based on its relevance and quality signals, not on whether it carries a speakable markup flag.

The four schema types with demonstrable AEO impact

The Organisation schema is the highest-leverage type for AEO because it establishes the entity that is the source of the information. When an AI system cites a piece of content, it needs to attribute it to an entity. Organisation schema with a complete description, url, and logo reference gives the AI system the raw material for that attribution. The critical fields are name, url, description, and sameAs (for known social profiles and Wikipedia entries). Missing or incomplete sameAs is the most common Organisation schema error that reduces its AEO utility.

Article schema on blog posts and news content is the second high-impact type. The Article schema with headline, datePublished, dateModified, author (pointing to a Person schema with a complete profile), publisher (pointing to the Organisation schema), and image produces consistent citation behaviour in AI search systems because it allows the AI to verify that the content is current, attributed, and from a credible source. The author field must point to a schema node with a complete Person definition, not just a name string. The AI systems that produce the most reliable citations are cross-referencing the author profile against the publisher organisation to verify legitimacy, and incomplete author markup is a signal that reduces citation confidence.

FAQPage schema with genuinely comprehensive FAQ content is the third high-impact type, but with a critical condition: the FAQ content must be substantive. FAQPage where each answer is two sentences of generic text will not produce citations. FAQPage where each answer is a 100 to 200 word substantive response to a real question does produce citations, because AI systems are now using depth of answer as a quality signal when evaluating FAQ markup. The practical implication is that if you are going to use FAQPage schema, you should write the FAQ answers as if they are standalone editorial content, not as markup padding.

HowTo schema is the fourth high-impact type for pages that genuinely contain step-by-step instructional content. HowTo with a complete step array, each step with a name and text (or an image with alt text for visual instructions), and a description field that contextualises the entire procedure, produces strong AEO citations for process-oriented queries. The condition for HowTo to work is that the page must contain the full procedure in the body content. HowTo markup on a page that has a summary but links to the full steps elsewhere does not produce citations because the AI system evaluates the content it can read, not the markup declaration.

The six schema types widely recommended for AEO that produce almost nothing

LocalBusiness schema is the most commonly overused type for AEO purposes. Most LocalBusiness implementations are incomplete, missing the critical fields that AI systems actually use (service area, opening hours accuracy, and the aggregate rating vs review count distinction). More importantly, for B2B content targeting commercial decision-makers, LocalBusiness schema is largely irrelevant because the queries that drive AEO citations in B2B contexts are not location-specific.

Review and AggregateRating schema produce almost no observable AEO impact for most content categories. The exception is product review content where the review is the primary content of the page and the aggregate rating represents a genuine consensus across a meaningful number of individual reviews. In practice, most review markup on B2B content sites is either too thin or too promotional to influence AI citation behaviour.

BreadcrumbList schema is decorative for AEO purposes. It provides no entity signal that AI systems use for citation decisions and is included in most AEO schema recommendations based on its value for traditional search navigation, not for AI search.

Event schema, JobPosting schema, and Recipe schema are all high-value for their specific search contexts but have essentially no impact on the AEO citation patterns for the content types most B2B operators produce, which are informational blog posts, service pages, and case studies.

Implementation: clean markup without the bloat

The clean approach to schema markup for AEO is to implement four node types completely rather than twelve node types partially. Organisation, Article, FAQPage (where content depth justifies it), and HowTo (where the content is a genuine procedure) cover the high-impact types. Every other schema type should justify its inclusion with a specific content match rather than with a generic “it might help” rationale.

The technical implementation requires JSON-LD in a script tag in the document head. The Organisation node should be in the document head as a self-contained graph node. The Article node should reference the Organisation node via the publisher field using the sameAs reference pattern that AI systems use for entity resolution. The FAQPage node should only be on pages where the FAQ content in the body matches the markup depth. The HowTo node should only be on pages that contain the full procedure in readable body content, not just in an introductory summary.

Validation should be run against the Google Rich Results Test and the Schema Markup Validator before deployment. The Google tool will confirm that the markup is syntactically valid and that the types that Google recognises are correctly formed. It will not confirm that the markup is useful for AEO, which requires evaluating whether the marked-up content actually matches the quality signals AI systems are looking for.

The most common schema markup for AEO implementation error is leaving dateModified off Article schema. datePublished is almost always present. dateModified is frequently omitted, which means the AI system cannot distinguish between an article published in 2021 and never updated and an article published in 2021 and comprehensively revised in 2026. For rapidly evolving topics in AI and technology, dateModified is not optional.

For the full AEO picture beyond schema markup, see our guide to answer engine optimisation for B2B services. For the foundational concepts behind AEO and how it differs from traditional SEO, see our earlier piece on what is AEO.

Schema markup for AEO is not a box to check. It is a quality signal layer that either accurately represents your content or misrepresents it, and AI systems are increasingly good at knowing the difference.

Frequently asked questions

Does every page on a B2B website need schema markup for AEO?

No. The highest-value approach is to implement complete, correct markup on your top twenty highest-traffic informational pages rather than incomplete markup on every page. AI systems evaluate pages individually, and thin markup on a high-traffic page damages your citation probability more than no markup at all. A single Article schema node with complete fields on a well-written post is worth more than ten half-complete Organisation and WebSite nodes across pages with generic content.

How does FAQPage schema interact with AI citation behaviour when the FAQ answers are short?

Short FAQ answers, typically anything under 50 words, do not produce reliable AEO citations because AI systems use content depth as a quality signal. FAQPage with 8 questions each answered in two sentences will be deprioritised relative to a single-question FAQPage where the answer is a 200-word substantive response. The practical implication is to use FAQPage on fewer pages with more depth per answer rather than on every page with minimal answer content.

Should AI-generated content use Article schema?

Article schema does not prevent AI content from being cited, but AI content that is thin, generic, or obviously templated will not be cited regardless of the markup. The schema markup communicates to AI systems that the content is a dated, attributed article from a known entity. Whether AI systems choose to cite it depends on whether the content is genuinely useful relative to other available sources on the same query. If you are producing AI-assisted content, Article schema is appropriate as long as the content quality is sufficient to justify the attribution signal the markup creates.

If you want a structured review of your current schema implementation against AEO best practice, speak to the AxionIQ team. We run a content technical audit as part of our AEO setup work that covers markup, content depth, and entity signal clarity.