Schema Markup Implementation: Before and After

Key Results

| Metric | Before | After | Change | |---|---|---|---| | AI Visibility Score | 28/100 | 63/100 | +35 points | | AI citations per week | 3 | 22 | +633% | | Pages interpreted correctly by AI | 16% | 54% | +238% improvement | | AI referral traffic (monthly) | 89 visits | 1,240 visits | +1,293% |

Company Overview

GreenLeaf Home (name changed for confidentiality) is a mid-sized e-commerce company selling sustainable home goods and eco-friendly products across Europe. With 320 product pages, a 45-article blog, and an FAQ knowledge base of 80 pages, the company had a substantial content footprint. Their Google SEO was solid -- they ranked on page 1 for 180+ keywords related to sustainable home products and eco-friendly alternatives.

But when the marketing team ran their first AI Visibility Score check in late 2025, the result was sobering: 28 out of 100. Despite having quality content and strong Google rankings, AI models were barely aware GreenLeaf Home existed.

The diagnosis was straightforward: the website had zero structured data. No JSON-LD markup of any kind. AI models that visited GreenLeaf Home's pages had to parse raw HTML and guess what the content was about -- a process that is unreliable at best and leads to misinterpretation or complete omission at worst. To understand why this matters at a fundamental level, see our guide on what AI SEO is.

The Challenge

The audit identified one primary problem with multiple downstream effects:

Zero structured data. GreenLeaf Home had no JSON-LD schema markup anywhere on the site. This meant:

• AI models could not identify GreenLeaf Home as a business entity -- there was no Organization schema to establish the company's name, type, location, or social profiles
• Product pages lacked Product and Offer schema, so AI had no structured way to understand pricing, availability, or product categories
• Blog articles had no Article schema, making it impossible for AI to assess content type, author, or publication date
• The 80 FAQ pages had no FAQ schema, despite containing exactly the kind of question-answer content that AI models preferentially cite

AI content interpretation test. The team ran a controlled test: they asked ChatGPT, Gemini, and Perplexity to describe GreenLeaf Home and its products. The results were telling:

• ChatGPT described the company as "a blog about sustainability" -- missing entirely that it was an e-commerce store
• Gemini returned no results for the brand name
• Perplexity cited one blog post but misattributed the company's product category

Without structured data to guide interpretation, AI models were guessing -- and guessing wrong.

The baseline. Before any changes, the team measured exactly where they stood using our JSON-LD basics for AI SEO framework:

| Schema Type | Status | Pages Affected | |---|---|---| | Organization | Missing | All (445 pages) | | Product + Offer | Missing | 320 product pages | | Article | Missing | 45 blog posts | | FAQPage | Missing | 80 FAQ pages | | BreadcrumbList | Missing | All (445 pages) |

The Strategy

The implementation followed a phased approach prioritized by impact and dependency:

Phase 1 (Week 1): Organization schema. Deploy Organization schema site-wide to establish the company entity. This is the foundation that all other schema builds upon -- AI models need to know who you are before they can correctly attribute your content.

Phase 2 (Weeks 2-3): FAQ schema. Add FAQPage schema to all 80 FAQ pages. This was prioritized second because FAQ content has the highest citation rate per page -- AI models can extract question-answer pairs directly.

Phase 3 (Weeks 3-5): Article schema. Deploy Article schema on all 45 blog posts with proper author, date, and section metadata.

Phase 4 (Weeks 5-8): Product schema. Add Product and Offer schema to all 320 product pages with category, pricing, availability, and aggregate ratings.

Implementation

Phase 1: Organization schema (Week 1)

The development team added Organization schema to the site-wide template, present on every page. The schema included:

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "GreenLeaf Home",
  "url": "https://greenleafhome.eu",
  "logo": "https://greenleafhome.eu/logo.png",
  "foundingDate": "2019",
  "description": "European e-commerce store specializing in sustainable home goods and eco-friendly products.",
  "sameAs": [
    "https://www.linkedin.com/company/greenleafhome",
    "https://www.instagram.com/greenleafhome",
    "https://www.facebook.com/greenleafhome"
  ],
  "contactPoint": {
    "@type": "ContactPoint",
    "telephone": "+48-xxx-xxx-xxx",
    "contactType": "customer service"
  }
}

Within 5 days of deploying Organization schema, the team re-tested AI responses. ChatGPT now correctly identified GreenLeaf Home as "a European e-commerce store for sustainable home products" -- no longer a "blog about sustainability." This single change fixed the fundamental entity recognition problem.

Phase 2: FAQ schema (Weeks 2-3)

The 80 FAQ pages received FAQPage schema with JSON-LD markup matching every visible question-answer pair. Each page typically contained 5-8 questions.

The team took an important extra step: they audited each FAQ answer for citability. Answers that were vague or too short were rewritten to be 50-100 words -- long enough to be a complete, standalone response, but concise enough for AI to extract as a citation.

Before optimization (typical FAQ answer):

"Yes, we ship to most European countries."

After optimization:

"GreenLeaf Home ships to 24 European countries including Germany, France, Netherlands, Spain, Italy, and all EU member states. Standard shipping takes 3-5 business days within the EU and 7-10 days for non-EU European countries. Free shipping is available on orders above 75 EUR."

The difference: the optimized answer contains specific, quotable facts that AI models can cite directly when users ask "Does GreenLeaf Home ship to Germany?" or "What is the shipping time for eco-friendly products in Europe?"

Phase 3: Article schema (Weeks 3-5)

All 45 blog posts received Article schema with:

• Headline and description
• Author information (linked to Person schema for the content team)
• datePublished and dateModified (corrected -- several articles had been updated but still showed original dates)
• articleSection categorization
• wordCount for content quality signaling

Phase 4: Product schema (Weeks 5-8)

The 320 product pages received Product and Offer schema. This was the most complex phase because product data needed to be dynamically pulled from the e-commerce platform:

• Product schema with name, description, brand, category, and image
• Offer schema with price, currency, availability, and shipping details
• AggregateRating schema pulled from customer reviews (average 4.3 stars across 12,000 reviews)

The dynamic schema ensured that price changes and stock updates were automatically reflected in the structured data.

Results

AI Visibility Score progression

The AI Visibility Score improved steadily with each phase of implementation:

| Phase | Week | AI Score | Change | Key Event | |---|---|---|---|---| | Baseline | 0 | 28 | -- | No schema | | Organization | 1 | 34 | +6 | Entity recognized | | FAQ schema | 3 | 45 | +11 | FAQ citations begin | | Article schema | 5 | 52 | +7 | Blog content discovered | | Product schema | 8 | 63 | +11 | Product recommendations appear |

Citation and traffic metrics

| Metric | Week 0 | Week 4 | Week 8 | |---|---|---|---| | AI citations/week | 3 | 11 | 22 | | Monthly AI referral visits | 89 | 480 | 1,240 | | Correct AI interpretation rate | 16% | 38% | 54% | | ChatGPT product recommendations | 0 | 3/week | 9/week | | Perplexity source citations | 2/week | 6/week | 11/week |

Business impact

The schema implementation drove measurable commercial results:

• AI-referred revenue grew from 230 EUR/month to 4,870 EUR/month by Week 8 -- a 21x increase
• Product recommendation citations led to an average order value of 94 EUR from AI referrals, 35% higher than the site average of 70 EUR
• Customer support ticket volume decreased 12% as AI models began citing FAQ answers directly, reducing the number of users who needed to contact support for routine questions
• Brand accuracy improved -- by Week 8, all three major AI platforms correctly described GreenLeaf Home's business, product categories, and shipping policies

Schema type impact ranking

When the team analyzed which schema type contributed most to the overall improvement:

1. Organization schema -- responsible for the foundational entity recognition fix. Without it, subsequent schema types would have been attributed to an unrecognized entity.
2. FAQ schema -- highest per-page citation impact. FAQ pages went from 0 citations to an average of 2.3 citations per page per week.
3. Product schema -- drove the most commercial value. Product recommendation citations had the highest conversion rate (3.8% vs 1.1% site average).
4. Article schema -- improved blog content discoverability and helped AI models assess content recency and authority.

Key Takeaways

1. Schema markup is the highest-ROI single intervention for AI SEO. A 35-point AI Score improvement from structured data alone demonstrates that schema is not optional -- it is foundational. Start with our JSON-LD basics for AI SEO guide.

2. Organization schema is the foundation. Deploy it first. AI models need to recognize your entity before they can correctly attribute your content. See our Organization schema authority guide for implementation details.

3. FAQ schema has the highest per-page impact. Pages with FAQ schema saw AI content interpretation jump from 16% to 54%. If you have FAQ content, adding FAQ schema should be your second priority after Organization.

4. Rewrite FAQ answers for citability. Short, vague answers are not useful to AI. Each answer should be 50-100 words, contain specific facts, and stand alone as a complete response.

5. Measure with AI Visibility Score. Use AI Score tracking to monitor progress phase by phase. Without measurement, you cannot attribute improvements to specific changes.