AI
Schema for AI Search: What Actually Matters (and What's Just Noise)
By Kyle Senger
15+ years in local marketing; Google Ads certified; Shopify Partner.
Here's something I notice when I audit SMB websites: most of them have zero schema markup. Some have a little. Almost none have it set up in a way that actually helps them show up in AI-generated answers.
And that's a real problem right now. Because schema for AI search isn't the same conversation it was two or three years ago. Back then, schema was mostly about rich snippets in Google, maybe a star rating under your listing. Now it's about whether ChatGPT, Perplexity, and Google's AI Overviews can confidently pull structured facts about your business and use them in an answer.
This article is specifically about schema markup and how it connects to AI search. I'm not going to cover the broader picture of AI in marketing here, because our complete guide to AI for marketing already does that well. What I'm going to do is get specific: which schema types actually matter, how to implement them without hiring a developer, and what the realistic payoff looks like for a Canadian SMB.
What Schema Markup Actually Does (Plain English Version)
Schema markup is code you add to your website that tells search engines and AI systems what your content means, not just what it says.
Without schema, a crawler reads your "About" page and sees words. With schema, it reads your page and understands: this is a local business, they're a dentist, they're located in Saskatoon, their hours are Monday to Friday 8am to 5pm, and their phone number is X. That's a fundamentally different level of clarity.
The technical format is called JSON-LD, and it lives in a <script> tag in your page's <head>. It doesn't change what visitors see. It's purely for machines.
Here's why that matters for AI search specifically. When ChatGPT or Perplexity pulls together an answer about "best accountants in Regina," it's drawing on structured data it's already indexed or retrieved. If your site has clean, accurate schema, you're giving those systems a confident signal. If your site has no schema, they're guessing, or worse, skipping you entirely.
I want to be clear: schema alone won't make you rank in AI answers. It's one piece of a larger puzzle. For the full picture on how AI systems surface content, see our breakdown of generative engine optimization and how to earn AI citations from ChatGPT and Perplexity. But schema is the foundation. Get it wrong and everything else you do gets harder.
The Schema Types That Actually Move the Needle for SMBs
There are hundreds of schema types on schema.org. Most of them don't matter for small and medium businesses. Here are the ones that do.
LocalBusiness (and its subtypes)
If you're a local business, this is non-negotiable. LocalBusiness schema tells AI systems your name, address, phone number, hours, and what type of business you are. The subtypes matter too: Dentist, LegalService, Plumber, AccountingService are all more specific than LocalBusiness and carry more signal weight.
What to include at minimum:
name(exactly as it appears on your Google Business Profile)addresswithstreetAddress,addressLocality,addressRegion,postalCode,addressCountrytelephoneopeningHoursSpecificationurlgeo(latitude and longitude , you can get these from Google Maps)
Consistency matters here. In my experience, when the name, address, and phone in your schema don't match your Google Business Profile exactly, AI systems get confused about which data to trust. That inconsistency shows up more often than you'd think, especially in businesses that have moved locations or changed their phone number.
FAQPage
This one is underused. FAQPage schema marks up question-and-answer content on your site so AI systems can pull it directly into responses.
Think about the questions your customers actually ask before they hire you. "How much does a dental cleaning cost in Winnipeg?" "Do I need a lawyer for a real estate transaction in Alberta?" "How long does HVAC installation take?" If you've written honest answers to those questions and marked them up with FAQPage schema, you're giving AI Overviews and Perplexity a clean, citable source.
The key word there is honest. Thin, vague answers don't get cited. Specific, useful answers do.
Article and BlogPosting
If you publish content, Article or BlogPosting schema helps AI systems understand authorship, publication date, and what the content is about. The author property is increasingly important because AI systems are paying more attention to E-E-A-T signals (that's Experience, Expertise, Authoritativeness, and Trustworthiness) when deciding what to cite.
Include author with a link to a Person schema that has your credentials, your name, and ideally a link to your About page. This is one of the quieter ways to build credibility with AI systems over time.
Review and AggregateRating
If you have reviews, mark them up. AggregateRating schema tells AI systems your average rating and how many reviews you have. This shows up in rich snippets in Google and signals credibility to AI systems pulling together comparisons.
One caveat: you can only mark up reviews that actually live on your website. You can't pull in your Google reviews and mark those up with schema. This is a common mistake I see. If you want this schema to work, you need a reviews section on your own site.
Service
For service businesses, Service schema lets you describe individual services with names, descriptions, and areas served. This is especially useful if you want to show up when someone asks an AI assistant "who does commercial HVAC in Calgary" or "which law firms in Toronto handle employment disputes."
How to Actually Implement Schema (Without a Developer)
Here's the honest process, week by week.
Week 1: Audit what you have.
Go to Google's Rich Results Test (search for it, it's free) and run your homepage and a few key service pages through it. This shows you what schema is currently on your site and whether it's valid. Most SMB sites come back with either nothing or errors. Write down what's missing.
Also check Search Console under Enhancements. If you have structured data errors, they'll show up there with enough detail to diagnose them.
Week 2: Build your LocalBusiness schema.
Use Google's Structured Data Markup Helper or a free JSON-LD generator (Schema.dev is a solid one). Fill in every field you can. Don't leave gaps. Copy the output and paste it into a <script type="application/ld+json"> tag in the <head> of your site.
If you're on WordPress, Rank Math or Yoast SEO will generate LocalBusiness schema for you through their settings. It's not perfect but it's a reasonable starting point. If you're on Squarespace or Wix, your options are more limited, and you may need to add it manually through a code injection field.
Week 3: Add FAQPage schema to your top three service pages.
Pick the pages that describe your most important services. Write three to five real questions with real answers on each page. Use the FAQ block in your CMS if it has one, then layer the schema on top. If you're doing it manually, the JSON-LD structure is straightforward: an FAQPage type with mainEntity as an array of Question items, each with acceptedAnswer.
Week 4: Validate everything and fix errors.
Run every page you touched through the Rich Results Test again. Fix any errors before moving on. Common mistakes: missing required fields, wrong property names, mismatched data between schema and visible page content (Google calls this a "mismatch" and it can get your structured data ignored or penalised).
Month 2: Add Article schema to your blog posts and Service schema to your service pages.
This is lower urgency than the above, but worth doing. If you publish regularly, set up a template in your CMS so new posts automatically get the right schema. Manual schema on every post is not a sustainable process.
The Canadian Specifics Worth Knowing
A few things that matter if you're operating in Canada.
Your address format. Schema expects addressRegion to be the two-letter province code (SK, AB, ON, BC, etc.) and addressCountry to be "CA". I've seen sites use full province names or "Canada" in the wrong field. It works, but the standardised format is cleaner.
Bilingual businesses. If you serve Quebec or have a bilingual site, you should implement schema in both languages where relevant. Quebec's Bill 96 requires French-language service in Quebec, and your schema should reflect the same language your page is in. Don't put English schema on a French page.
PIPEDA and what you're marking up. Schema markup is about your business information, not customer data, so PIPEDA doesn't really come into play here. But if you're marking up anything that touches individual users (like review authors), keep it to publicly available information.
Local service areas. The areaServed property in your LocalBusiness schema lets you specify where you operate. For a plumber in Saskatoon who also serves Martensville and Warman, that's worth including. AI systems use this to answer "who does [service] near [location]" queries, and being explicit about your service area helps.
What the Realistic Payoff Looks Like
I want to be straight with you here, because I've seen agencies oversell schema as some kind of magic.
Schema markup is infrastructure. It makes everything else you do more effective. It's not a shortcut to ranking in AI answers overnight.
Per DataForSEO data, "AI search optimization" in Canada gets around 140 searches per month at a CPC of CA$17.72. That's a signal that businesses are actively spending money trying to solve this problem. Schema is one of the cheapest and most durable pieces of that solution.
Across sites I've worked on, the pattern I see is this: businesses that have clean, complete schema tend to show up more consistently in AI-generated local answers than businesses with messy or missing schema. Not always. But consistently enough that it's worth the few hours it takes to get right.
Here's a rough worked example. Say you're a trades company in Edmonton running Google Ads at CA$3,000/month (ad spend only). Your cost per lead from paid is around CA$85 based on a 3.5% conversion rate on a landing page getting 1,000 clicks a month. That math gives you roughly 35 leads at CA$85 each. If your schema is clean and you start picking up even five additional organic leads per month from AI search, you're adding leads at effectively CA$0 marginal cost. Five leads at your average job value of, say, CA$1,200 is CA$6,000 in pipeline from a one-time setup that took a few hours.
I'm not promising those numbers. I'm showing you the shape of the opportunity. Your actual numbers will depend on your market, your competition, and how well the rest of your site is set up.
For a full technical audit of how AI systems are reading your site right now, our AI SEO audit guide walks through the exact process. And if you want to understand how schema fits into the broader question of ranking in AI-generated answers specifically, our guide to ranking in Google AI Overviews covers the content side of that equation.
FAQ: Schema for AI Search
Does schema markup directly influence ChatGPT or Perplexity?
Not in the same way it influences Google. ChatGPT and Perplexity don't read your schema in real time the way Google's crawler does. But schema makes your content cleaner and more structured, which means when those systems do index or retrieve your content, the information is clearer and more citable. Think of it as reducing ambiguity rather than sending a direct signal.
How often do I need to update my schema?
Whenever something changes: new address, new hours, new phone number, new services. Stale schema is worse than no schema in some cases, because it creates conflicting signals. Build a habit of checking your schema any time you update your Google Business Profile.
My site is on Wix or Squarespace. Can I still add schema?
Yes, but with more friction. Both platforms have code injection fields where you can paste JSON-LD. Wix also has some built-in structured data for certain site types. It's not as clean as a WordPress setup with a proper SEO plugin, but it works. If you're on a platform that genuinely doesn't allow custom code, that's a real limitation worth factoring into your next site decision.
Is there a risk of getting penalised for schema?
Yes, if you mark up content that isn't actually on the page. Google calls this "spammy structured data" and it can result in manual actions. The rule is simple: only mark up what a visitor can actually see on the page. Don't add AggregateRating schema if there are no visible reviews. Don't add FAQPage schema if there's no visible FAQ section.
What about AI-specific schema types, like speakable?
Speakable schema was designed for voice assistants and was in beta for a long time. It's worth knowing about but not a priority for most SMBs right now. The foundational types I described above (LocalBusiness, FAQPage, Article) will do more for you. Our LLM SEO guide covers some of the more forward-looking optimisation approaches if you want to go deeper.
3 Takeaways
Schema is infrastructure, not a shortcut. It makes your site easier for AI systems to understand and cite. It won't fix bad content or a weak Google Business Profile. It will make everything else you do more effective.
Start with LocalBusiness and FAQPage. Those two schema types do the most work for the most SMBs. Everything else is useful but secondary. Don't let perfect be the enemy of done.
Consistency matters more than completeness. Your schema, your Google Business Profile, and your website should all say the same thing about your business. Conflicts between those sources create confusion for AI systems and can cost you citations you'd otherwise earn.
