Why you need to get nerdy and learn about schema markup

I started building Frilly as a way to find out what events are happening near me in the Madison area. But then, the AI bug bit me, and I found myself going down the rabbit hole of AI discovery more broadly. The more I researched, the more it seemed that the problem of incomplete or nonexistent technical SEO on websites was exacerbating the problem of discovery in the LLM-enhanced future.

Think about what happens when someone asks a chatbot, “What’s a good bakery in north Madison?”

The AI accesses bakery websites and evaluates the results to answer the question. It looks for structured data. These are machine-readable signals that say “this is a bakery, it’s at this address, here’s what makes it special.” If that structured data doesn’t exist, a bakery that could become my favorite is unlikely to be mentioned.

This is the problem schema markup solves, and it’s the technical foundation of every Frilly profile.

Comprehensive coverage for every entity

My goal is for Frilly to become exceedingly good at creating the intelligence layer for local brands and their existing websites. To that end, the platform is designed so that every business, organization, and individual maker in the Frilly directory gets a complete Schema.org JSON-LD markup. This is the same structured data format that Google, Microsoft, and the major AI platforms use to understand the web.

Here’s what that means in practice:

For Ancora Cafe + Bakery, a Madison staple on Sherman Avenue, we generate markup that tells AI systems:

• This is a LocalBusiness
• Its canonical identity is https://frilly.ai/b/ancora-cafe-bakery
• It’s located at 611 North Sherman Avenue in Madison, WI 53704
• Its geo-coordinates are 43.1077°N, 89.3630°W, and
• Here’s a rich description of what makes it special

But we go further. We link that schema to two AI-optimized documents (a markdown profile and a structured JSON file) that give AI systems deeper, richer context than a schema snippet alone could provide.

We’re just getting started

I recently watched a webinar by Schema App, a company that helps enterprise brands optimize their structured data for AI search. They laid out a maturity matrix for schema markup. The dimensions are things like: are you using the right schema types? How deep are your properties? Is your markup covering the whole site or just a few pages? Are your entities connected to each other? It was super informative.

So, being honest about where we are:

Following that rubric, I was happy to see that Frilly sits solidly at the Intermediate level across most of these dimensions, and we are genuinely Advanced in one area: coverage. Every single one of the 1,000+ entities and 600+ upcoming events in our directory gets the same comprehensive markup treatment. That’s not how most sites work. Up until now, it’s been laborious to place schema markup on anything other than the homepage and a few product pages. We do it for every profile, consistently.

Where we’re headed next is the Advanced tier. As we fill the directory with more local brands, we’ll be:

• Adding more precise schema types (using CafeOrCoffeeShop instead of LocalBusiness for a cafe)
• Connecting entities to each other through their markup (so a neighborhood’s schema references the businesses in it), and
• Linking to external knowledge bases like Wikidata and Google’s Knowledge Graph.

The infrastructure is already there — every entity has a stable @id URI. We just need to wire the connections!

The intelligence layer that sets this local directory apart

The schema markup is just one layer. On top of it, we’ve built what I think of as the AEO (Answer Engine Optimization) infrastructure. This is a set of systems specifically designed for how AI actually discovers and processes information. We’ve also added the content negotiation protocol that serves an AI agent (like Claude) the markdown file of your profile instead of the token-hungry HTML. This is important to note because Cloudflare, a web infrastructure company, has recently embraced this paradigm.

Every Frilly profile has a direct markdown URL (like this) that is served in the content negotiation phase and a JSON URL (like this ) that AI systems can fetch without parsing any HTML.

When someone asks an AI assistant about local businesses in Madison, these are the signals that increase the chances of a brand showing up in the answer. The other layer is the quality of your content, and that is something we could help with through a direct engagement, if you’re interested. But the technical part of it is free.

Why this matters for your local brand

The brands that will be visible in the new world where AI agents can navigate a parallel web aren’t necessarily the ones with the biggest ad budgets. They’re the ones whose information is structured, accurate, and machine-readable.

The way I think about it: most local business websites are optimized for human visitors, and the product designer in me loves that. Beautiful photos, compelling copy, easy-to-navigate menus.

But AI systems don’t browse websites the way people do. They need structured data, semantic markup, machine-readable formats. Most local businesses don’t have the resources or technical know-how to add that layer themselves.

That’s where Frilly comes in. We’re adding the intelligence layer to your existing web presence. Our goal is to make your brand visible and understandable to AI through these profiles through the Smart Directory and the Smart Chat. And if you’d like help upgrading your own site beyond that, reach out, we’d love to talk.

We’re actively tracking the standards as they evolve — from Schema.org property updates to new AI discovery protocols — and applying them to your profiles automatically.

That’s the thesis: local knowledge, structured for the AI era, maintained by people who care about the community it represents.

Claim yours and help make it better

Look for your brand in the Smart Directory, and if you can’t find yourself, join here!

Your participation will help us create a map of the local ecosystem that leads to better discovery over time. I’ll be sharing details on our progress with more numbers and nerdy details.