Lawrence Hitches Written by Lawrence Hitches | AI SEO Consultant | May 05, 2026 | 7 min read

Schema.org Action types let you declare to AI agents and search engines what actions are available on your page, search the catalog, buy this product, reserve this table, subscribe to this list. Google has supported them since 2014, but most SEOs have never implemented them. In 2026, with AI agents (Claude Computer Use, OpenAI Operator, Microsoft Copilot Vision) increasingly executing tasks on user behalf, Action schema becomes the cleanest way to tell agents "here's exactly what you can do here." Pages with Action schema get surfaced in agent task-completion flows; pages without rely on the agent inferring intent from prose. This guide covers the Action types Google actually supports, how to implement them, the practical SearchAction + ReserveAction + BuyAction patterns, and how Action schema interacts with the emerging WebMCP standard.

What Are Schema.org Action Types?

Schema.org has hundreds of types. Most SEOs use a small subset: Article, Product, LocalBusiness, Organization, FAQPage, BreadcrumbList. Action is a different beast, instead of describing what something is, Action describes what someone can do.

The Action vocabulary covers actions a user can perform on or with the entity described. Examples:

  • SearchAction, search this site or content
  • BuyAction, purchase this product
  • ReserveAction, reserve this resource (table, room, appointment)
  • RegisterAction, register for an event or service
  • SubscribeAction, subscribe to a newsletter or service
  • ListenAction, play this audio
  • WatchAction, play this video
  • ReadAction, read this content
  • DonateAction, donate to this cause
  • OrderAction, place an order

Each Action takes properties: target (the URL or endpoint that handles the action), actionStatus (whether it's available), potentialAction on the entity it relates to.

Why Action Schema Matters for AI Agents in 2026

Three reasons SEO teams should care now:

  1. Agents read schema as the structural ground truth. When Claude Computer Use, OpenAI Operator, or Microsoft Copilot Vision parses a page to figure out what actions are available, well-formed Action schema is the cleanest signal. Inferring from button text and form fields is messier and more error-prone.
  2. Google's SearchAction unlocks the sitelink search box, the search box that appears under your homepage in Google's SERP for branded queries. Direct traffic + better brand experience. Almost every brand site qualifies; few implement it.
  3. Action schema is the bridge to WebMCP. When the WebMCP standard ships (2026-2027), the actions declared there will look very similar to Schema.org Action types. Sites with Action schema today are halfway to WebMCP. See our WebMCP guide.

The 3 Action Types Worth Implementing First

1. SearchAction (universal, easiest win)

Lets Google show a search box directly under your homepage in the SERP. Also tells AI agents exactly how to query your site's search.

{
  "@context": "https://schema.org",
  "@type": "WebSite",
  "url": "/",
  "potentialAction": {
    "@type": "SearchAction",
    "target": {
      "@type": "EntryPoint",
      "urlTemplate": "/search/?q={search_term_string}"
    },
    "query-input": "required name=search_term_string"
  }
}

Drop in your homepage. Google may surface a sitelink search box on branded queries. Agents can call your search endpoint directly without parsing the search UI.

2. ReserveAction (booking businesses, restaurants, services)

For any business that takes bookings, restaurants, salons, consultations, services. Tells agents exactly how to initiate a reservation.

{
  "@context": "https://schema.org",
  "@type": "Service",
  "name": "AI SEO Consultation",
  "provider": {"@type": "Person", "name": "Lawrence Hitches"},
  "potentialAction": {
    "@type": "ReserveAction",
    "target": {
      "@type": "EntryPoint",
      "urlTemplate": "/contact/?topic={topic}",
      "actionPlatform": ["http://schema.org/DesktopWebPlatform", "http://schema.org/MobileWebPlatform"]
    },
    "result": {
      "@type": "Reservation",
      "name": "AI SEO Consultation Booking"
    }
  }
}

Agents booking on user behalf can call this URL with the appropriate parameters. Without it, the agent has to navigate to your contact page, parse the form, infer field meanings, fill in plausible values, hit submit, much higher failure rate.

3. BuyAction (product pages on ecommerce)

Product schema with embedded BuyAction tells agents exactly how to add to cart and complete a purchase flow.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Example Product",
  "offers": {
    "@type": "Offer",
    "price": "49.99",
    "priceCurrency": "AUD",
    "availability": "https://schema.org/InStock",
    "potentialAction": {
      "@type": "BuyAction",
      "target": {
        "@type": "EntryPoint",
        "urlTemplate": "https://example.com/cart/add?sku={sku}&qty={quantity}"
      }
    }
  },
  "sku": "EXAMPLE-001"
}

Agents executing "buy this product" on user behalf can hit the cart endpoint directly. Massive improvement over scraping the product page, finding the buy button, clicking it, parsing the cart confirmation.

Action Types Google Actually Supports for Rich Results

Google's documentation specifies which Action types unlock visible SERP enhancements:

Action typeGoogle rich resultStatus (2026)
SearchActionSitelink search boxActive, widely available
WatchActionVideo carousel resultsLimited; Video schema preferred
ListenActionPodcast carouselsLimited; PodcastSeries schema preferred
ReadActionReading list featuresLimited rollout
ReserveActionReserve buttons in some verticals (limited)Limited; mostly used for restaurant/event integrations
BuyActionBuy buttons (limited verticals)Limited; merchant integration required

The visible SERP impact of Action schema is modest. The bigger payoff is agent action accuracy, which Google's docs don't directly call out but which Anthropic, OpenAI, and Microsoft documentation increasingly does.

How Agents Actually Use Action Schema

Pattern across the major agent implementations in 2026:

  1. Claude Computer Use, parses the DOM and accessibility tree, looks for declared schema, prefers calling Action endpoints over inferring UI clicks. Documented improvement when target sites have Action schema.
  2. OpenAI Operator, similar pattern; uses both Schema.org Action types and (when available) any WebMCP manifest. Operator-friendly sites in early Operator demos all had Action schema.
  3. Microsoft Copilot Vision, more vision-heavy, uses schema as a fallback when vision interpretation is ambiguous. Still meaningful boost from Action schema for transactional flows.
  4. Google Gemini agents (in development), Google has been clear that Gemini agentic features will lean on existing structured data heavily, including Action types. Sites with comprehensive schema will be the canonical sources.

Action Schema vs WebMCP: How They Relate

Schema.org Action types and WebMCP overlap conceptually but serve different deployment models:

AspectSchema.org ActionWebMCP
Standard ageSince 2011 (mature)Early preview April 2026
FormatJSON-LD inside the page <head>Separate manifest file (proposed)
AudienceSearch engines + AI engines + agentsAI agents primarily
Action expressivenessPer-page declarativeSite-wide capability declaration with parameters and outcomes
Adoption status (2026)Widely supported, lightly usedEarly adopter program only

Practical: implement Schema.org Action types now. They work today across all major search engines and agents. When WebMCP stabilises, you'll likely run both, Action schema for per-page declarations + WebMCP for site-wide manifest of all available actions. Coverage of WebMCP at our WebMCP guide.

Common Action Schema Mistakes

From auditing client implementations:

  1. Action without target. The whole point is the URL endpoint that handles the action. Without target, the schema is decorative.
  2. Hardcoded URLs without templates. Use urlTemplate with parameter placeholders (e.g., {search_term_string}) so agents can substitute values.
  3. Wrong Action type. A "request a quote" form is not a BuyAction, it's likely a ContactAction or a custom Action subtype. Match the type to the actual user intent.
  4. SearchAction pointing to a non-functional URL. Test the target URL, paste a sample query, confirm the search results page loads with that query.
  5. Multiple SearchActions on the same page. One per WebSite entity is enough.
  6. Missing actionPlatform for booking actions. Helps agents understand whether the action works on desktop, mobile, or both.
  7. Putting Action schema on the wrong entity. SearchAction goes on the WebSite entity (homepage). BuyAction goes on the Product entity (product page). ReserveAction goes on the Service or Restaurant entity.

How to Test Action Schema

  1. Google Rich Results Test, validates the JSON-LD syntax and confirms which rich result enhancements your Action qualifies for: search.google.com/test/rich-results
  2. Schema.org Validator, generic spec validation: validator.schema.org
  3. Manual agent test, ask Claude (with Computer Use) or OpenAI Operator to perform the action on your site. Note whether it uses your declared endpoint vs. clicking through the UI. The former is the agent finding and using your schema.

FAQ: Schema.org Action Types for AI Agents

What are Schema.org Action types?

Action types are Schema.org vocabulary for declaring what users can do on a page or with an entity. Examples: SearchAction (search this site), BuyAction (purchase this product), ReserveAction (reserve a table or appointment), RegisterAction (register for an event), SubscribeAction (subscribe to a service). Each Action takes a target URL endpoint that handles the action.

Does Schema.org Action affect SEO rankings?

Indirectly. SearchAction unlocks Google's sitelink search box for branded queries (visible SERP enhancement). Other Action types have limited visible SERP impact but increasingly serve AI agents (Claude Computer Use, OpenAI Operator, Microsoft Copilot Vision) parsing pages to determine available actions. The bigger 2026 payoff is agent action accuracy, not direct ranking.

How is Schema.org Action different from WebMCP?

Schema.org Action is mature (since 2011), declared per-page in JSON-LD, and supported by all major search engines and agents today. WebMCP is in early preview (April 2026), uses a separate site-wide manifest file, and targets AI agents specifically. They overlap conceptually but serve different deployment models. Practical: implement Action schema now; layer WebMCP when it stabilises.

Which Action type should I implement first?

SearchAction. It's the only one with reliably visible SERP impact (sitelink search box on Google for branded queries) and works on essentially any site with a search function. Drop the JSON-LD on your homepage, point the target URL at your search results endpoint with a query parameter, done.

Does Google actually use BuyAction and ReserveAction?

Limited rich result rollout in 2026. Google supports the schema but visible SERP buy/reserve buttons are restricted to specific verticals with merchant integrations. The bigger value is agent comprehension, Anthropic, OpenAI, and Microsoft agent documentation increasingly references Action schema as a primary structural signal for transactional pages.

How do I validate Action schema?

Google Rich Results Test (validates syntax + rich result eligibility) and Schema.org Validator (generic spec validation). Both free, instant feedback. For agent-specific validation, manually test by asking Claude Computer Use or OpenAI Operator to perform the action on your site, observe whether the agent calls your declared endpoint or falls back to clicking through the UI.

What's the most common Action schema mistake?

Implementing the schema without a functional target URL. The whole point is the endpoint that handles the action. Without a real target URL (and ideally a urlTemplate with parameter placeholders), the schema is decorative, technically valid but functionally useless for agents.

Sources & Further Reading

Keep Reading

Soaring Above Search

Weekly AI search insights from the front line. One newsletter. Six sections. Everything that actually moved this week, with a practitioner's take.

Lawrence Hitches
Lawrence Hitches AI SEO Consultant, Melbourne

Chief of Staff at StudioHawk, Australia's largest dedicated SEO agency. Specialising in AI search visibility, technical SEO, and organic growth strategy. Leading a team of 120+ across Melbourne, Sydney, London, and the US. Book a free consultation →