There are CMOs that are spending millions on content while their competitors appear on ChatGPT, Perplexity and Claude with half the budget.
The difference isn't the quality of the content. The thing is that one brand has structured data and the other doesn't.
When a user asks Chat GPT “best CRM for small teams”, the model doesn't read your entire website. Look for data that you can extract quickly. Price, features, integrations, user limits.
If that information is in prose paragraphs, the model ignores it (because it's expensive to crawl). If it's in schema markup, it processes it in milliseconds.
Structured data went from being “nice to have for rich snippets” to critical infrastructure to exist in search of AI.
Traditional search engines crawl, index, rank. AI models extract, synthesize, cite.
Google reads your page and decides if it deserves to rank for a query. ChatGPT reads your page and decides if it has data to use to respond. They are different processes.
A software customer had full product pages. Detailed descriptions, use cases, pricing. But the pricing was written like this: “Our basic plan starts at just $29 a month, perfect for freelancers and small businesses.”
When someone asked ChatGPT “how much does [product name] cost”, Perplexity didn't cite their site. I was quoting a competitor who had the price in Product schema markup with price: 29, PriceCurrency: USD.
Not that customer content was worse. The thing is that it was harder to extract.
Not all types of schema have the same impact. Some are decorative. Others are functional.
If you sell something, this isn't optional. Price, availability, rating, reviews. AI models prioritize sites that have this clean data.
An e-commerce customer implemented Product schema on 2,400 SKUs. Three months later, appearances in ChatGPT responses had risen 35%. They didn't change the content. They only structured what already existed.
FAQs are the perfect format for AI models. Specific question, concise answer. Exactly what they need to quote.
But it has to be marked. An FAQ written in normal HTML is just text. An FAQ with schema tells the model “this is a question, this is the answer, use this.”
An insurance customer had 40 FAQs on his site. Well written, useful. Few appearances in AI. We added FAQ schema. Two weeks later, they began to appear as a cited source for 12 of those questions.
If your content includes processes, steps, instructions, this schema is critical. Models use it to give step-by-step answers.
I haven't really tried this one with clients in production, but if you have an informational blog or content where the steps make sense and you share processes, you should include it.
Basic information about your company. Name, logo, contact, locations. It seems trivial, but models use it to verify that you're a real company before they quote you.
If you don't have this, the model may doubt if your site is legitimate. Especially in industries such as finance or healthcare where credibility matters.
The models cite sites with social proof. If you have 500 5-star reviews but they're not in the diagram, the model doesn't see them. If they are marked, they become part of your authority.
This one is my favorite because it is the one with which I have seen incredible results. A client in a highly regulated industry had G2 reviews embedded on his site. But without a pattern.
We added AggregateRating with RatingValue: 4.7 and ReviewCount: 487 (the real one, you don't want to lie with this). A little over a month later, they began to appear in LLM responses as “one of the highest-rated options.”
Google used structured data primarily for rich snippets. If your schema was wrong, you would lose the snippet but you were still ranking.
AI models use structured data to decide whether to cite you. If your schema is wrong or doesn't exist, you don't show up. Dot.
In case the examples I have given you so far are not enough for you, I'll give you another one. A customer had content about interest rates. Good content, well researched. But the rates were in HTML tables without a schema.
Your competitor had the same rates in a table with schema markup Table. When you asked ChatGPT for “interest rates for [account type]”, the competitor appeared 89% of the time. My client, 3%.
It wasn't domain authority. It wasn't backlinks. It was extraction. We changed it and although we haven't yet achieved that 89% of citations, we achieved 21%.
Most companies have poorly implemented or incomplete structured data.
Common errors:
They implemented Organization schema in 2019 and never updated it. The logo changed, the address changed, the phone number changed. The diagram is still showing old data.
That's why I'm so reluctant to implement schema on sites that don't have technical capacity and where a request to IT takes 5 to 8 business months.
They mark the homepage but the product pages, blog posts, service pages, nothing. Models don't just read your homepage, they read your site in 360.
They have a Product schema but the price is missing. Or they have a FAQ schema but only in 5 of 40 FAQs. Half-scale implementation does not generate full results.
Google Search Console shows schema errors that no one checks. If Google can't read your schema, neither can AI models.
The temptation is to bookmark the entire site at once. Bad idea. Mass deployment without strategy generates errors, wastes time, and dilutes effort.
Most sites already have some schema. The problem is that no one knows exactly what, where, or if it's working.
Use Google Search Console. Go to Enhancements. It shows you what types of schema Google detects and how many errors you have. If you have 200 errors in Product schema, fix that before adding a new schema.
Also use Schema.org Schema Markup Validator. Copy the HTML of your main pages and check what you find. Often you discover an old schema that was left by a developer years ago and that is generating silent errors.
Not all pages deserve the same effort. Start where it matters.
JSON-LD is the format that Google recommends. It's easier to implement, easier to maintain, and doesn't interfere with your HTML.
It goes in the <head>or at the end of the<body>. You don't have to touch your visible content. This means that you can add, edit, or remove schema without breaking the design of your site.
Microdata and RDFa require direct HTML markup. If your designer changes the structure, the schema breaks. With JSON-LD, they're separate.
If you have 10 pages, mark them by hand. If you have 1,000, you need automation.
Schema with errors is worse than having no schema. Google ignores it, AI models ignore it, and you're leaving useless code on your site. I have really seen horrible results from having poorly implemented structured data.
Again, I repeat, if you don't have any technical skills, it's better not to have them. If you have a product and you change the price on your site but not in the schema, not only can you be misrepresented in answers, but in some cases you can even perform much worse than if you had nothing.
The schema is not static. If you change the price of a product, update the schema. If you add a new question to your FAQ, add it to the schema.
If you're using automation, the update should be automatic. If you marked manually, create a process. Every time you update content, check if the schema needs updating as well.
Once the first schemas are working and validated, expand.
If you started with Product schema on your top 10 products, add it to the next 20. If you started with FAQ schema in 5 questions, mark them all.
But do it in phases. Implement, validate, measure, expand. Don't skip validation and measurement.
Two years ago, structured data was advanced optimization. Today they are a basic requirement.
AI models aren't going to start reading dense prose to extract prices, features, or instructions. They will continue to prioritize sites where that data is marked and ready to consume.
Your competitors are already implementing it. Some good, many bad, but they're doing it. The window where this was a differentiator is closing. Soon it will be the entry price.
If you don't have structured data yet, start this week. If you already have them, check that they are correct and complete. If they are correct, measure if they are generating appearances in AI.
.png){kind=small}
SEO consulting that scales traffic, grows your brand, increases revenue, and helps you win on Google and AI platforms and LLMs like ChatGPT.
%402x.svg){kind=icon}
