Schema markup and AI visibility: the no-nonsense technical guide
Which schemas to implement, in what order, and why? Technical guide based on our audits to maximize your visibility in LLM responses.
| Schema | Priority | AI Impact | Difficulty |
|---|---|---|---|
Organization | 1 | High | Easy |
FAQPage | 2 | High | Easy |
Article / BlogPosting | 3 | Medium-high | Easy |
HowTo | 4 | Medium | Medium |
Product / Service | 5 | Medium | Medium |
LocalBusiness | 6 | Medium (local) | Easy |
BreadcrumbList | 7 | Low-medium | Easy |
This ranking was built after 50+ technical audits. Not by reading the Schema.org docs for an afternoon. We correlated the presence of each schema type with citation frequency in AI responses. The result is clear: some schemas matter a lot, others are noise.
Organization: your identity card for machines
If you implement only one schema, make it this one. The Organization schema tells crawlers and LLMs: here's who we are, where we are, what we do, and where to find us elsewhere on the web.
Here are the fields that really matter:
name: your exact business name. Not a variant, not an acronym. The name as you want to be identified.
description: one sentence. Clear. No marketing jargon.
url: your main website.
sameAs: an array with your LinkedIn, Google Business, Crunchbase profiles, etc. This is where you link all your presences together.
address: your physical address. LLMs use geolocation heavily.
foundingDate, founders: optional but useful. They add depth to your entity.
We often see sites with an Organization schema that only contains the name and URL. That's like sending a resume with just your first name. Technically correct, practically useless.
FAQPage: the most direct entry point
LLMs love FAQs. Why? Because the question-answer format corresponds exactly to how they work. A user asks a question, the model looks for an answer. If your site already has the question AND the answer in a structured format, you're making the model's job easier.
The key: don't put just any questions. Put the questions people actually ask LLMs in your sector. How to find them? Use Perplexity, type queries related to your activity, and observe the "related questions" that appear. Those are your FAQs.
Technically, the FAQPage schema is simple. An array of Question each with an acceptedAnswer. No limit on number, but we recommend 5 to 10 per page. Beyond that, Google may ignore the schema and LLMs see it as structured keyword stuffing.
Article and BlogPosting: structure your expertise
Every blog article should have its Article or BlogPosting schema. It's basic, but we find that 60% of sites we audit either don't have it or have it poorly implemented.
The important fields beyond title and date:
author: with a link to an author page. LLMs evaluate author credibility. An article signed "Admin" carries less weight than one signed by a person with a verifiable LinkedIn profile.
datePublished and dateModified: LLMs favor recent content. If you update an article, change the modification date.
speakable: rarely used, but useful for voice assistants. Identifies passages that can be "spoken" as an answer.
HowTo: for procedural content
If you have "how to" articles, the HowTo schema is a significant bonus. It structures each step with a name, description, and optionally an image or required tool.
We tested this on a client in the construction sector in Namur. Same article, with and without HowTo schema. With the schema, Perplexity cited the article by listing the steps directly in its response. Without the schema, the article appeared as just another link. The difference between being the answer and being a source.
Product and Service: say what you sell
When someone asks an LLM "what tool for managing invoices in Belgium?", the model looks for Product or Service entities with clear attributes: name, description, price, geographic area, category.
If your service page doesn't have this schema, you're relying on the LLM to "understand" that you sell something by reading your marketing prose. Good luck. Service pages are often the worst structured: lots of storytelling, little machine-exploitable data.
Our advice: keep the storytelling for humans, add the schema for machines. Both can coexist on the same page.
The implementation order
If you're starting from zero, here's how we proceed with our clients:
Week 1: Organization on the homepage. It's the foundation. 30 minutes of work for a developer.
Week 2: FAQPage on your 3 to 5 most visited pages. Identify the questions, write concise answers, add the schema.
Week 3: Article/BlogPosting on all existing articles. If you have a CMS like WordPress, a plugin like Yoast or RankMath does this automatically. Check that the data is correct anyway.
Week 4: Product/Service on your commercial pages and HowTo on your guides.
Total: one month. No redesign needed. No agency charging $15K. A competent developer and someone who knows the business is enough.
Mistakes that undo all the work
Invisible schema. The JSON-LD is in the code, but it contains syntax errors. Always validate with the Rich Results Test from Google and the Schema.org validator.
Schema that doesn't match visible content. Your FAQ schema lists 8 questions, but only 3 are visible on the page. Google calls this "schema spam" and may penalize. LLMs will simply ignore the signal.
Empty or generic fields. "description": "We are an innovative company". That says nothing. Be specific: sector, location, concrete offering.
Schema markup isn't magic. It's a structured language that machines read natively. If you speak it correctly, they understand you. If you botch it, they ignore you. It's that simple.
