Most service businesses either skip schema entirely or paste in a generic LocalBusiness block and call it done. Neither approach does much for you in 2026, when AI assistants are actively parsing structured data to decide what to tell customers about who you are, what you offer, and whether to recommend you. The schema types that matter are specific — and a handful of them do most of the work.
Why schema matters more now than it did for Google
Google's crawler has always been good at reading plain text. AI assistants are less reliable at it. When ChatGPT or Perplexity reads your website, it's working from a snapshot: compressed, often stripped of layout context, sometimes truncated. Schema markup is machine-readable metadata that travels cleanly through that process. A plumber's site with no schema might get a vague "they do plumbing stuff" summary. The same site with proper markup gives an AI enough to say "Riverside Pipe & Drain handles water heater replacement, slab leak detection, and emergency calls — open until 8 p.m. weekdays."
That specificity is the difference between a referral and a shrug.
LocalBusiness with a typed subclass
The LocalBusiness type is the foundation, but the subclass is what sharpens the signal. Schema.org has specific types for Plumber, Dentist, HVACBusiness, LegalService, AutoRepair, BeautySalon, and dozens more. Using the right subclass tells AI systems — not just Google — exactly what category of business you are without leaving them to infer it.
Include at minimum: name, address (with PostalAddress sub-properties), telephone, url, openingHoursSpecification, and areaServed. The areaServed property is particularly underused — it lets you list the cities or regions you cover, which is how an AI learns to recommend you for "plumber near Westbrook" even if your office is in a neighboring town.
Service and OfferCatalog
This is the most important schema type most service businesses are missing. Service lets you describe individual offerings — not just "we do HVAC" but each distinct service: furnace tune-up, AC installation, duct cleaning, heat pump repair. Wrap multiple Service entries in an OfferCatalog and you give AI assistants a structured list they can actually cite.
A dental practice, for example, can list Service items for teeth whitening, Invisalign, emergency extractions, and routine cleanings — each with its own name, description, and optionally a PriceSpecification. AI assistants that answer "what does Maplewood Family Dental offer?" will pull from exactly this data if it exists. Without it, they guess from page text, and guesses are often incomplete or wrong.
You can see how LocalBusiness schema applies to this kind of structured setup in the full implementation guide.
FAQPage
FAQPage schema is direct fuel for AI-generated answers. Each Question and Answer pair you mark up becomes a ready-made response that AI assistants can quote, paraphrase, or link to. For service businesses, the highest-value FAQ items are:
- What does a service cost (even a range)?
- How quickly can you respond or book an appointment?
- What areas do you serve?
- What should a customer do in an emergency situation?
- Are you licensed, bonded, or certified for this work?
A roofing contractor who marks up "Do you offer free estimates?" with a clear answer is directly shaping what ChatGPT says when someone asks about getting a quote. A contractor without that markup is relying on the AI to figure it out — which it often doesn't.
AggregateRating
Reviews already show up on Google. But AggregateRating schema makes your review count and score machine-readable to AI systems that might not cross-reference your Google Business Profile at all. Include ratingValue, reviewCount, and bestRating. A family law attorney sitting at 4.8 stars across 200 reviews has a strong signal — but only if that data is marked up where AI can read it, not just where Google can index it.
This is one of the factors that helps AI assistants recommend you when someone asks who to call. Social proof needs to be structured, not just visible.
OpeningHoursSpecification (done correctly)
The plain openingHours string property ("Mo-Fr 09:00-18:00") works for Google but is easy for AI systems to misread or ignore. OpeningHoursSpecification as a full sub-object — with dayOfWeek, opens, and closes — is the more reliable format. Specify holiday closures with validFrom and validThrough. If you have seasonal hours, add a second OpeningHoursSpecification block covering that period.
Wrong hours are one of the most common AI errors for service businesses. You can read more about why AI gets business information wrong and what to do about it — hours are consistently near the top of the list.
PriceSpecification inside Service
You don't need to publish a fixed price for every job. But a range or starting price inside a PriceSpecification block on each Service entry does real work: it answers the question customers most want answered before they call, and it gives AI assistants a credible data point instead of "pricing varies — contact for a quote," which tells no one anything.
A house cleaning service that marks up "$120–$200 for a standard two-bedroom clean" will appear more useful in AI responses than a competitor whose site says "affordable rates tailored to your needs."
How to check what you have
Run your homepage and key service pages through Google's Rich Results Test, but also check the raw JSON-LD output manually — AI systems read the raw markup, not just what Google validates. Look for missing areaServed, absent Service items, and openingHours in plain string format rather than the full specification object.
A free scan will show you which of these signals AI assistants are currently reading correctly for your business — and where the gaps are that change what customers hear when they ask.