How We Structure Schema Markup for AI Overview Visibility
Written by Gabriel Bertolo
April 2, 2026
Most websites that have the schema markup box checked move on. They installed Yoast or Rank Math, let the plugin auto-generate some basic structured data, and never thought about it again. The schema says the page is an “Article.” It says the business is a “LocalBusiness.” Maybe it pulls in the site name and a logo.
That’s not schema markup. That’s the bare minimum. And AI engines treat it that way.
I write the schema by hand. Not with a plugin. Not with an AI generator. I personally write entity-rich JSON-LD for every client I work with, because the depth of what I can communicate to Google AI Overviews, ChatGPT, Perplexity, and other AI search engines through custom structured data is fundamentally different from what any automated tool produces.
This isn’t a philosophical preference. I’ve seen the results. A Boston-based therapy practice with zero structured data went from hemorrhaging traffic to a 45.2% increase in organic clicks after we implemented advanced Person Schema, LocalBusiness Schema, and entity-rich markup across the site. A B2B manufacturer with a Domain Rating of 21 achieved the #1 position in Google AI Overview ahead of FDA.gov and two Fortune 100 competitors, and the entity-rich schema connecting their brand to regulatory terms and industry certifications was a core part of how we got there.
This article is my actual process. How I think about schema to increase AI search visibility, which types matter most, how I connect them into an entity graph, and what I’ve seen work based on 13+ years of implementing this for real businesses.
Let me be specific about what plugins actually generate. When Yoast or Rank Math creates schema for a blog post, you get something like this: an Article type with a headline, a datePublished, and maybe an author name as a plain text string. For your business, you get a generic Organization type with your site name and URL.
That’s it. No credentials for the author. No knowsAbout declarations telling AI what topics this person is an expert in. No sameAs links connecting your business to its LinkedIn profile, Google Business Profile, industry directories, or any other external validation. No @id references linking the article’s author to the Organization they work for. No connection between Service offerings and the business entity. Each schema block sits in isolation on its page, disconnected from everything else.
AI engines don’t just check whether schema exists. They evaluate its depth, specificity, and interconnectedness. A standalone Article schema has marginal value. But an Article schema connected via @id to a Person schema (with real credentials, a jobTitle, knowsAbout properties, and sameAs links to external profiles) connected to an Organization schema (with location, services, reviews, and its own sameAs connections) creates a semantic graph that AI engines can traverse. That’s a fundamentally different signal.
Only about 31% of websites implement any schema markup at all. Source – Almcorp
And of that 31%, the vast majority are using plugin defaults. The competitive bar is genuinely low. If you implement schema with real depth, you’re immediately in a different category than 90%+ of your competitors.
This isn’t theoretical either. In March 2025, both Google and Microsoft publicly confirmed that they use structured data for their generative AI features. Google was explicit: structured data is critical for modern search features because it is efficient, precise, and easy for machines to process. ChatGPT confirmed it uses structured data to determine which content appears in its results. The platforms that decide whether your business gets cited in AI-generated responses are telling you directly that schema matters.
There are over 800 schema types on Schema.org. Most of them are irrelevant to what we’re trying to accomplish. Based on what I’ve implemented across healthcare, legal, professional services, e-commerce, and B2B clients, here are the types that consistently produce results.
This is the foundation. It establishes your business as a distinct entity in the knowledge graph. Your name, URL, logo, address, phone number, and social/directory links via the sameAs property.
The sameAs property is where most implementations fall short. It’s not optional for AI visibility. AI systems cross-reference entities across multiple sources to build confidence. When your Organization schema includes sameAs links to your LinkedIn company page, your Google Business Profile, your Yelp listing, your industry directory profiles, and any other authoritative external presence, you’re giving AI engines a network of verification points. Each one says: this entity exists in multiple trusted places, and they all agree on who it is.
This is the one most businesses completely miss. And for YMYL industries like healthcare, legal, and financial services, it might be the single most impactful schema type you can implement.
For the Boston therapy practice, writing Person Schema for each therapist was a central factor in the recovery. Each practitioner got a full entity profile: name, jobTitle, worksFor (linked via @id to the Organization), credentials, specializations, degrees, professional affiliations, knowsAbout (declaring their areas of clinical expertise), and sameAs links to their professional profiles. Google finally had machine-readable proof that real, qualified experts were behind the content.
Before the schema implementation, Google had no way to verify the therapists’ expertise. After it, Google could see: this article about anxiety treatment was written by a licensed clinical psychologist with 20 years of experience who specializes in cognitive behavioral therapy and works for this specific practice. That’s a completely different trust signal than an anonymous article on a website with no author credentials.
AI platforms weigh content authored by recognized entities higher than anonymous content. This is the authorship signal that builds E-E-A-T at the structured data level, and for YMYL sites it’s not optional.
Don’t use generic LocalBusiness. Use the most specific subtype that applies: MedicalBusiness for healthcare practices, LegalService for law firms, ProfessionalService for consultants, Restaurant for restaurants. AI engines use these subtypes to match your business with the right intent categories. A personal injury firm and a family law mediator should not be described the same way in schema.
Include geo coordinates, service area, opening hours, accepted payment methods, and aggregate ratings. And align every detail exactly with your Google Business Profile data. When schema and GBP contradict each other, even on something small like business hours, AI systems don’t average the signals. They discount the schema and sometimes ignore it entirely.
FAQPage schema shows the highest citation probability among all schema types for AI search. Source – WPRiders
This makes sense when you think about how LLMs work. They respond to questions with synthesized answers. FAQPage schema is literally a set of questions and answers in a format AI engines can directly extract. Each FAQ becomes both a potential rich result in traditional Google search AND an AI extraction point.
I implement FAQPage schema on service pages and in blog content. On a service page, the FAQs answer the questions a prospect would ask before hiring: “How much does this cost?” “How long does it take?” “What’s included?” In blog content, the FAQs address the broader informational queries that drive traffic. Both give AI engines structured, citation-ready content.
Every piece of content on your site needs to be connected to its author and publisher through proper schema references. The Article schema should include the headline, datePublished, dateModified (AI engines favor recently updated content), and about properties declaring the main topics.
But the critical piece is the author and publisher connections. The author property should point to the Person’s @id. The publisher property should point to the Organization’s @id. This connects the content to the credentialed human who wrote it and the verified business that published it. Without these connections, the article is just text on a page with no verifiable authority behind it.
For service-based businesses, this fills a gap that Product schema can’t. AI engines need to understand what you offer, where you offer it, and under which entity. Use serviceType, provider (linked to your Organization via @id), and areaServed. If you have multiple service categories, use hasOfferCatalog to organize them. When someone asks an AI “What SEO services are available in Boston?”, the structured data you’ve defined is what helps the AI determine whether to include you.
This is where the real differentiation happens, and it’s the part that no plugin handles.
Most schema implementations treat each block as an island. One page has Article schema. Another page has Organization schema. A team page might have Person schema. They’re all separate. AI engines can’t connect them because nothing tells the AI that these entities are related.
My approach: every entity gets a stable @id. Think of it like a permanent internal address for each entity in your schema.
The Organization gets an @id like https://www.radiantelephant.com/#organization. My personal Person schema gets an @id like https://www.radiantelephant.com/#gabriel-bertolo. These @ids are consistent across every page of the site.
Then every other schema block on the site references these @ids instead of duplicating information. When I write Article schema for a blog post, the author property doesn’t just say “Gabriel Bertolo” as a text string. It points to my Person @id. The publisher property points to the Organization @id. My Person schema has a worksFor property pointing back to the Organization @id. The Organization has a founder property pointing to my Person @id. Each Service schema has a provider property pointing to the Organization @id.
The result is a connected graph. Not isolated data fragments. AI engines can traverse these connections and build a complete picture: this article was written by this person, who has these credentials and expertise areas, who founded this organization, which offers these services, in these locations, and has these reviews and external validations.
A recent Search Engine Land analysis put it well: in traditional SEO, many implementations stop at adding Article or Organization markup in isolation. For AI search, the useful pattern is to connect nodes into a coherent graph using @id. That connected graph is what I build for every client.
I want to be clear about something. Schema alone doesn’t produce rankings. It’s one layer within a broader SEO and Generative Engine Optimization strategy. But without it, AI engines don’t have the structured context to understand who your business is, what you do, and why you should be trusted. Here’s how it played out in two very different client situations.
This practice employed some of the most qualified therapists in the region. Licensed professionals with decades of clinical experience. But Google had zero structured signals to verify any of it. No Person Schema. No author credentials. No entity connections.
I wrote advanced Person Schema for each therapist on the team. Full credential profiles: degrees, licenses, specializations, professional affiliations, knowsAbout declarations for their clinical expertise areas, and sameAs links to their professional profiles. I connected each Person entity via @id to the Organization, and connected each article to its author.
This was part of a comprehensive technical SEO recovery that also included metadata overhaul, internal linking restructuring, and backlink cleanup. But the schema was the piece that gave Google machine-readable proof of the expertise that was always there but invisible to algorithms.
The results over 90 days: organic traffic grew 45.2%. Search impressions increased 95.3%. Click-through rate improved 235%, from 1.29% to 4.32%.
Individual pages jumped 18 positions. For a YMYL healthcare site, the Person Schema was the difference between Google guessing about author credibility and Google knowing.
Different situation entirely. A multi-vertical B2B manufacturer with a Domain Rating of 21 competing against companies with DRs of 60-80, including FDA.gov and Fortune 100 brands like Cardinal Health and McKesson.
The schema strategy here was about entity relationships at scale. I built structured data connecting the client’s brand to regulatory terms, industry certifications, product categories, and manufacturing capabilities across five verticals. The goal was to make the client’s entity so clearly defined and so deeply connected to the relevant industry knowledge that AI engines would recognize them as an authoritative source despite the lower domain authority.
Combined with a broader GEO strategy including content development and off-page authority building, the client achieved #1 position in Google AI Overview for their primary industry queries. Domain Rating went from 21 to 35 in seven months. Branded searches increased 587%.
A DR 21 site outranking FDA.gov in AI search shouldn’t happen based on traditional SEO logic. But in AI search, the entity clarity and structured context that schema provides can override raw domain authority. That’s the reality I’ve observed.
I audit a lot of sites. These are the problems I find most frequently, and each one is actively holding back AI visibility.
Plugin defaults left untouched. The schema says “Article” and “Organization” with no meaningful entity data. Name and URL. That’s it. AI engines treat this as background noise.
Schema contradicts on-page content. Business hours in schema don’t match the Google Business Profile. Review ratings in schema don’t match visible reviews on the page. When schema contradicts visible content, Google doesn’t try to reconcile the difference. It ignores the schema altogether. I’ve seen sites with otherwise good SEO lose rich results entirely because of these inconsistencies.
No Person Schema for authors or team members. Especially damaging for YMYL industries. If your therapists, lawyers, doctors, or financial advisors aren’t represented in Person Schema with their actual credentials, AI has no structured way to evaluate their expertise. Their content gets treated the same as anonymous content.
Generic types instead of specific subtypes. Using “LocalBusiness” when “MedicalBusiness” or “LegalService” would be more accurate. The specific subtypes help AI match your business to the right intent categories. Precision matters.
Schema implemented once and never updated. Team members leave. Services change. Hours shift. Prices change. When schema doesn’t reflect current reality, the inconsistencies compound over time and erode the trust signals you’ve built.
Isolated blocks with no @id connections. Organization schema on the homepage. Article schema on blog posts. Person schema on the team page. None of them linked to each other. AI sees fragments instead of a coherent entity.
If you want to see where your site stands, here’s a quick diagnostic.
Run your homepage and two or three key pages through the Google Rich Results Test (search.google.com/test/rich-results). This shows you exactly what structured data Google sees on each page, any errors, and which rich results you’re eligible for.
Check Google Search Console under the Enhancements section. It aggregates schema issues across your entire site and flags errors that are preventing rich results.
Then look at what’s actually in the schema. Is it just the plugin default? Does it include real credentials for your team? Are the @id references connecting entities across pages? Does it match your Google Business Profile data?
If you want to prioritize what to fix first, here’s the order I’d follow. Organization schema first to establish your core business entity. Person Schema for key team members next, especially if you’re in a YMYL industry. Then FAQPage schema on your highest-traffic pages. Then Article/BlogPosting with proper author and publisher @id connections. Then Service or Product schema depending on your business model.
Implementing comprehensive schema increases the likelihood of appearing in AI citations by up to 40%. Source – Search Engine Land
That stat aligns with what I’ve observed in practice. Schema doesn’t guarantee AI citations. You still need authoritative content, proper E-E-A-T signals, and topical relevance. But without structured data, even exceptional content faces an uphill battle because AI engines have to work harder to understand it, and they’ll often cite a competitor whose content is easier to parse.
I could use plugins. It would be faster. But the depth of schema I write for clients isn’t achievable with automated tools.
For some clients, I’m writing 100+ lines of JSON-LD per page. Declaring knowsAbout properties for each team member’s specific expertise areas. Building sameAs connections that link the business to every verified external presence. Connecting every entity in the graph through stable @id references. Writing long-form entity descriptions that give AI engines rich semantic context beyond just a name and URL.
Plugins can’t do this because they don’t know your business. They don’t know that your lead therapist specializes in trauma-informed CBT for first responders. They don’t know that your manufacturing facility holds three specific industry certifications. They don’t know that your law firm’s founding partner has been published in four legal journals. That specificity is what makes schema work for AI visibility, and it has to come from someone who understands the business.
At Radiant Elephant, schema markup isn’t a technical afterthought. It’s a core part of every SEO and GEO engagement. I treat it as the translation layer between your expertise and how machines understand it. The businesses that get cited in AI search are the ones where that translation is precise, thorough, and connected.
If you want to see what your current schema looks like and where the gaps are, schedule a free strategy call. I’ll run the audit and show you exactly what’s missing.
Gabriel Bertolo is a 3rd generation entrepreneur who founded Radiant Elephant over 13 years ago after working for various advertising and marketing agencies.
He is also an award-winning Jazz/Funk drummer and composer, as well as a visual artist.
His Web Design, SEO, and Marketing insights have been quoted in Forbes, Business Insider, Hubspot, Entrepreneur, Shopify, MECLABS, and more.
Check out some publications he's been quoted in:
Quoted in HubSpot's AI Search Visibility Article and HubSpot's Article on 6 Best Wix Alternatives
Quoted in DesignRush Dental Marketing Guide
Quoted in MECLABS
Quoted in DataBox Website Optimization Article and DataBox Best SEO Blogs
Quoted in Seoptimer
Quoted in Shopify Blog