Can we hand AI a client’s business details, add a list of services, and expect good schema markup? Not with that step alone. AI may produce code that passes a validator. But passing a test does not mean the schema connects the business, its services, its locations, and its website pages in the right way.
That is the part we need to get right. Our goal is not just valid schema. Our goal is connected schema that reflects the whole website.
We can use AI to help build that structure. But we need to give it more than a pile of facts and a request for code. We need a site crawl, a plan for how the data connects, and clear rules for each page type.
Valid Schema Is Only Part of the Job
When we ask an AI tool to create schema for one page, it may do exactly that. It takes the details we provide and builds a block of markup for that page.
The problem is what it may leave out: the links between that page’s data and the data on the rest of the site.
We can end up with separate pieces of schema that each pass validation but do not work together. A location page may describe a local business without properly referencing the related services or other relevant pages.
That gives us two separate checks:
- Does the markup validate? This checks whether the code meets the rules tested by the validator.
- Does the data connect? This checks whether related pages and entities reference one another as planned.
We want both. A valid block of code is a start, not the whole job.
Start With an On-Page Crawl
Before we generate schema, we crawl the website. In our workflow, we use the DataForSEO on-page crawl to gather the site’s pages and produce a crawl report.
That report becomes an input for our local schema generator. It gives the AI the sitemap, page URLs, and site information it needs before it starts writing markup.
This changes the task. Instead of asking AI to describe one page with limited context, we give it a view of the whole site. It can then plan how the pages and their data should fit together.
The order matters:
- Crawl the website.
- Collect the crawl report and page URLs.
- Map the relationships across the site.
- Generate schema for each page using that map.
The crawl is a starting point for schema, not a step we add after the code is done. We want the site structure in place before the first page’s markup is built.
Got SEO Questions? Get answers every week at 4pm ET at Hump Day Hangouts. Ask questions ahead of time, or live – just go to: https://semanticmastery.com/hdho (bookmark this!) 10+ years of insights given every week!
Get your checklist to help get better results with GBPs, faster.
Need DFY services for you Local SEO projects? Head over to Semantic Links.
Map the Graph Before Building Each Page
A schema graph is a way to connect related data. The term may sound complex, but the idea is simple. Our business, services, locations, and pages should not sit in separate boxes with no links between them.
A node is an item in that graph. Node references connect those items. With proper references, the schema can show how a page relates to the business, a service, or a location elsewhere in the site’s data.
We plan those connections before generating individual page markup. The AI uses the full site map to work out how the graph should fit together, then starts building the page-level schema.
This helps us avoid two problems:
- Disconnected data on one page: Schema items appear together but lack the right links between them.
- Disconnected data across pages: Each page has markup, but related items on other pages are not properly referenced.
We do not solve the second problem by fixing only the first. Our plan needs to cover the website, not just one page at a time.
Match the Schema to the Page’s Purpose
Once we have the graph planned, we can build schema for the homepage, contact page, about page, service pages, and location pages. We do not treat them all as if they serve the same purpose.
Service Pages
For a service page, our approach uses WebPage schema with Service as the main entity. That separates the page itself from the service the page describes.
The service still needs the right links to the rest of the graph. We are not just labeling the page and moving on. We are placing that service within the site’s wider structure, including service catalog relationships where they apply.
Location Pages
For a location page, we use WebPage schema with the appropriate LocalBusiness entity as the main entity. We use a more specific business subtype when it fits.
The location page should also have the proper references to relevant pages and related data. The exact links depend on the business and the site, which is another reason to crawl first.
Other Useful Properties
Our schema workflow can also include properties such as:
- about: What a page or item is about.
- knowsAbout: Topics associated with the business.
- areaServed: The areas the business serves.
- sameAs: Links that identify the same entity elsewhere.
We provide the relevant details, including the sameAs links. AI still needs the right inputs. A better workflow does not remove that need.
Use a Purpose-Built AI Skill
A broad prompt asks AI to work out the process from scratch. A purpose-built skill gives it a defined process to follow.
Our Local Schema Generator skill began with an existing schema skill. We added our own knowledge and changes, with a focus on local business schema. Although it can support other schema work, those additions were built around local businesses.
We use it in Cursor, and it can also be added to Codex Cloud. The tool matters less than the process: the schema skill receives the crawl data, plans the graph, and builds the markup with connected references.
For MasterMIND members, the Local Schema Generator is in the AI and Automation area. The related crawl skill is part of the Content Pipeline workflow. These work together so the crawl report becomes a prerequisite for schema generation.
A Better Brief for AI Schema Generation
We can keep the process clear without reducing it to a single vague prompt. Our brief should give AI the business facts and the context needed to connect them.
- The business name, address, and phone number.
- The services and locations represented on the site.
- The crawl report, sitemap, and page URLs.
- The relevant identity links for sameAs.
- The rules for page types and graph references.
Then we build from the site plan down to each page, rather than trying to stitch separate outputs together later.
AI can help us create good schema, but business details alone are not enough. We get a stronger result when we crawl first, map the relationships, and generate markup that stays connected across the site. That is the difference between code that merely passes a check and structured data built as one clear model.

