AEO Schema

Source: github.com/psyduckler/aeo-skills Part of: AEO Skills Suite (v2 Core) Previously: aeo-schema-optimizer — renamed in v2 for namespace consistency

Generate structured data (JSON-LD) optimized for AI citation. Help Gemini 3 Flash and other AI models find, understand, and cite your content.

Why Structured Data Matters for AEO

When Gemini 3 Flash generates AI Overviews, it needs to quickly identify:

• What type of content is on the page (article, FAQ, how-to guide, product)
• What questions it answers
• What entities it covers (people, products, organizations)
• How authoritative the source is (author, publication date, organization)

Schema markup explicitly declares all of this. Pages with proper schema give AI models a structured signal layer on top of the content — making them easier to parse, trust, and cite.

Requirements

• web_fetch — to analyze the target page
• LLM reasoning — to determine optimal schema types and generate JSON-LD
• Templates in references/schema-templates.md

No API keys required.

Input

• URL (required) — the page to optimize
• Target prompt (optional) — the AI prompt this page should win citations for

Workflow

Step 1: Fetch and Analyze the Page

Use web_fetch on the target URL. Extract:

1. Page title and meta description
2. Content structure — headings (H1-H3), sections, lists
3. Content type — determine what kind of page this is:
   • Article/Blog — long-form content, guides, analysis
   • FAQ — question-and-answer format
   • How-To — step-by-step instructions
   • Product — product page with specs, pricing, reviews
   • Local Business — location with address, hours, contact
   • Landing Page — mixed content, multiple purposes
4. Key entities — people (authors), organizations, products, tools mentioned
5. Existing schema — check for any <script type="application/ld+json"> blocks

Step 2: Audit Existing Schema

If schema already exists:

• Is it valid JSON-LD?
• Does it match the page content type?
• Is it complete (all recommended properties filled)?
• Are there missing schema types that would help? (e.g., article page without FAQ schema for Q&A sections)
• Are dates current (datePublished, dateModified)?

Output audit findings: what's present, what's missing, what's incorrect.

Step 3: Determine Optimal Schema Strategy

Based on content type and structure, recommend which schema types to implement:

Key principle: Layer multiple schema types. An article that includes a FAQ section should have both Article AND FAQPage schema. This gives AI models multiple structured entry points.

Step 4: Generate Optimized JSON-LD

Use templates from references/schema-templates.md and populate with actual page data.

Critical properties for AI citation:

For Article:

• headline — match the H1, keep under 110 chars
• description — the meta description or a 150-char summary
• author — full name + credentials URL (sameAs)
• datePublished + dateModified — ISO 8601, must be current
• publisher — organization with logo
• mainEntityOfPage — canonical URL
• about — what the article covers (link to entities)

For FAQPage:

• Extract every Q&A pair from the page
• Each Question gets the exact heading text
• Each acceptedAnswer gets the first 1-2 sentences (the direct answer)
• Limit to 10 most important Q&As

For HowTo:

• name — the process being described
• step — each step with name and text
• totalTime — ISO 8601 duration if mentioned
• estimatedCost — if applicable
• supply / tool — materials or tools needed

For Product:

• name, description, brand
• offers — pricing, availability
• aggregateRating — if reviews exist
• review — individual reviews

For LocalBusiness:

• name, address, telephone
• openingHoursSpecification
• geo — latitude/longitude
• priceRange

Step 5: Validate and Deliver

Before delivering the JSON-LD:

1. Validate JSON syntax — ensure it parses correctly
2. Check required properties — every schema type has required fields
3. Verify entity consistency — author, publisher, dates match across schema blocks
4. Test nested structures — FAQ questions, HowTo steps are properly structured
5. Check URL references — mainEntityOfPage, sameAs, url fields are correct

Output:

• Complete JSON-LD block(s) ready to paste into <head>
• Implementation notes (where to place, what to update when content changes)
• If existing schema was found: diff showing what changed

Step 6: Verification Guidance

After implementation, recommend:

• Test with Google Rich Results Test
• Validate with Schema.org Validator
• Wait 2-4 weeks for Google to re-crawl and process
• Use aeo-ai-overview-simulator to measure citation impact before/after

Schema Types Gemini Prioritizes

Based on observed AI Overview behavior, these schema signals matter most:

1. FAQPage — directly provides Q&A pairs the AI can cite; highest impact for question prompts
2. Article with dateModified — freshness signal; AI models prefer recently-updated content
3. HowTo — structured steps are easy for AI to extract and present
4. Author with credentials — expertise signal (E-E-A-T); helps for YMYL topics
5. BreadcrumbList — helps AI understand site structure and topic hierarchy
6. Product with reviews — for commercial queries; aggregateRating is a trust signal

Tips

• Don't over-schema. Only add schema types that match actual page content. Misleading schema can trigger penalties.
• FAQ is the highest-ROI schema for AEO. If your page answers questions (even implicitly), extract those Q&As into FAQPage schema.
• Keep dateModified current. Every time you update content, update the dateModified. Stale dates signal stale content.
• Author schema matters for YMYL. Health, finance, legal content benefits enormously from detailed author schema with credentials.
• Pair with content optimization. Schema helps AI find your answers; the answers themselves still need to be clear, concise, and extractable. Use aeo-content-free for that.