Published March 19, 2026
Business owners assume AI sees their website the way a customer does. It does not.
A customer visits a homepage and sees a logo, photos, a phone number in the footer, and a list of services. AI skips most of that. It does not see the logo, interpret photos, or browse like a human. AI reads the page looking for structured data that answers specific questions: what is this business, where is it, how can a user reach it, and is it available right now.
If those answers are in the schema, AI reads them directly. If they are not, AI tries to infer them from page content — which is slower, less reliable, and often incomplete. This gap is the reason most businesses are invisible to AI recommendations.
Most business owners believe that because their website looks professional and contains all the relevant information, AI understands their business. The phone number is on every page. The hours are in the footer. The services are listed with descriptions and photos.
That assumption is wrong. AI does not scan visually, interpret layout, or read a phone number in a footer image. It does not infer hours from a "Mon–Fri 8–5" string in a sidebar widget. Schema markup is how a website communicates with AI in AI's own language. When schema is present, AI reads it directly — field by field, with no ambiguity. When schema is absent, AI falls back to parsing unstructured page content, which is where things break down.
A typical homepage without schema: AI may pick up the business name from the title tag. Beyond that, it is guessing — no structured phone number, no hours, no service area, no specific services. Every piece of information not in schema is information AI has to infer, and inference is where errors happen.
The same homepage with complete schema: AI reads the JSON-LD and immediately knows the business type, name, address, phone number, email, hours, service area, and what the business does. No inference required.
The difference is not subtle. LLM visibility depends on giving AI structured answers to the questions it needs resolved. When those answers are missing, the business is missing from AI recommendations.
During a controlled experiment, a user followed up via email — to an address that does not appear anywhere on the website. Not in the footer, not on the contact page, not visible to any human visitor.
That email address exists in exactly one place: the Organization schema in the site's JSON-LD. The most likely explanation is that ChatGPT read the email directly from the schema and returned it to the user.
The implication is significant. The information in your JSON-LD is not just metadata that helps AI decide whether to recommend you — it may be the information AI provides directly to users. Your schema fields are front-line communication between your business and people who ask AI about you. If your schema contains outdated contact information, that is what AI may give to users. If it is missing, AI has nothing to provide.
"What is this business?" AI looks for business type in schema. Without it, a commercial roofing company categorized as "organization" competes against every other organization instead of against other contractors.
"Can the user reach them?" AI looks for phone, email, and address. As the schema email finding shows, AI may serve this information directly to users.
"Are they available right now?" AI looks for hours of operation. If hours are not in schema or in consistent third-party directories, AI cannot confirm availability.
"Do they match what the user needs?" Service descriptions, pricing, and service area let AI match the business to the specific query. A user asking for emergency plumbing in a specific city needs a plumber who serves that city and offers emergency services — not just any plumber.
Every field that is present moves the business closer to citation. Every field that is missing is a reason to recommend someone whose data is clearer to the system.
Run the AIFDS validator. Enter your URL at validator.aifds.org and see a field-by-field report of what is present and what is missing.
Compare what you think is there to what is actually there. Most business owners are surprised. The phone number in the footer is not the same as a phone number in JSON-LD. Hours in a sidebar widget are not the same as hours in structured data.
Send the results to your developer. The validator output tells your developer exactly which fields need to be added. The AIFDS blueprint library provides the exact JSON-LD for every industry and page type.
Think of your schema as a message to AI. The information in your JSON-LD may be what AI gives directly to people asking about your business. Make sure it is accurate, complete, and up to date — the same way you would keep your Google Business Profile current.
David Valencia writes about how AI systems find, parse, and cite websites.
The field-level findings behind what AI looks for.
Read →Why existing validators miss what AI needs.
Read →