Google Business Agent: The 48‑Hour Shopify Playbook (UCP, Endpoints, Testing, and Attribution)

Turn on Google’s Business Agent in 48 hours. This Shopify playbook covers UCP, endpoints, schema, QA queries, and revenue attribution—with copy‑paste snippets.

Google’s Business Agent is now live in Search, giving brands a native AI sales associate customers can engage with before they ever hit your site. If you’re on Shopify (or a similar stack), you can stand this up fast by aligning three things: clean product data, task‑ready endpoints, and airtight attribution.

This guide builds on our earlier explainers and kits. If you want the high‑level why and what, start here: Google’s Business Agent Is Live. For UCP fundamentals, read UCP Is Live: The 72‑Hour Ecommerce Fix and the Retailer’s UCP Starter Kit. Measurement? Don’t skip Assistant Attribution 101.

Who this is for

Shopify or Shopify Plus brands with 100–50,000 SKUs.
Teams that can ship small code changes (Liquid, theme app extensions) and expose simple JSON endpoints.
Founders and growth leads who need measurable revenue impact within a week.

Outcomes in 48 hours

Business Agent can answer accurate product questions, surface in‑stock variants, and hand customers a buy link.
Your catalog is agent‑ready via structured data and lightweight UCP‑style endpoints.
Every assistant‑driven click is tagged so you can see sessions, carts, and revenue in your analytics.

Prerequisites (Day 0)

Clean product data in Shopify: titles, variant options, prices, GTIN/MPN where available, shipping rules.
Merchant Center product feed or equivalent source of truth kept in sync with Shopify.
Search Console verified domain, XML sitemap live.
Legal pages visible and consistent (returns, warranty, shipping). Price shown = price charged.

Step 1 — Define intents, guardrails, and voice

Business Agent works best when you tell it exactly what it can do. Write one page that covers:

Top 10 intents: “Find size 9 trail running shoes under $120,” “Compare Model X vs Model Y,” “Is this hypoallergenic?”
Allow/deny list: No medical or safety claims beyond what’s on PDPs; do not negotiate price; always show disclaimer on availability.
Response style: 2–3 sentence answers + bullets; link to PDP or pre‑cart page with parameters.

Step 2 — Expose task‑ready endpoints (UCP‑style)

You don’t need a giant API to be agent‑ready. Start with three read endpoints and one action link:

2.1 Product lookup

{
  "GET /api/agent/products": {
    "q": "trail running shoes size 9 waterproof",
    "filters": {"price_max":120,"in_stock":true},
    "returns": [
      {"id":"sku_123","title":"Mesa Trail 2","price":109.00,"url":"/products/mesa-trail-2"}
    ]
  }
}

2.2 Availability and price

{
  "GET /api/agent/availability": {
    "sku": "sku_123",
    "returns": {"in_stock": true, "quantity": 42, "price": 109.00, "currency":"USD"}
  }
}

2.3 FAQs/attributes

{
  "GET /api/agent/attributes": {
    "sku": "sku_123",
    "returns": {
      "materials": "recycled mesh upper",
      "waterproof": true,
      "care": "hand wash, air dry"
    }
  }
}

2.4 Buy link (action)

Instead of taking payment off‑domain on day one, hand off to a pre‑cart page with pinned UTM tags:

https://yourstore.com/cart/add?variant=VARIANT_ID&quantity=1&utm_source=business_agent&utm_medium=organic&utm_campaign=assistant_buy

Tip: Implement these as a tiny edge function (Cloudflare/Netlify) that reads Shopify Storefront API and caches responses for 60–300 seconds. Keep responses under ~2 KB for snappy answers.

Step 3 — Tighten your product schema

Add complete Product + Offer JSON‑LD to PDPs. Keep the price and availability fields exactly synced with your buy link and availability endpoint.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Mesa Trail 2",
  "sku": "sku_123",
  "gtin13": "1234567890123",
  "brand": {"@type":"Brand","name":"TerraRun"},
  "description": "Lightweight waterproof trail runner.",
  "image": ["https://yourstore.com/images/mesa-trail-2.jpg"],
  "offers": {
    "@type": "Offer",
    "price": "109.00",
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock",
    "url": "https://yourstore.com/products/mesa-trail-2"
  }
}
</script>

Also ensure your collection pages and search results are crawlable and include clear headings, filters, and internal links.

Step 4 — Hand the agent a compact knowledge base

Create a single, crawlable page that aggregates concise specs and policies referenced in answers:

/kb/assistant — 800–1,200 words total. Sections: Sizing, Materials, Care, Shipping, Returns, Warranty.
Link each section to authoritative PDP snippets to keep answers grounded.
Update whenever policies change; keep language consistent with PDPs.

Step 5 — Compose answers that convert

Give the agent patterns that are short, helpful, and point to action:

Comparison: “Here are two waterproof trail shoes under $120. Option A grips mud; Option B is lighter for dry terrain. Prefer grip or speed?”
Objection: “It’s waterproof to 30 minutes of rain; for full-day storms consider Model Z.”
Action: “Ready to check out? Here’s a pre‑filled cart with size 9 in stock.”

Step 6 — QA with high‑intent queries

Before you call it done, run targeted “agent QA” queries. You’re testing for three things: accuracy, actionability, and attribution.

“Show women’s black leggings with pockets under $60 in size M.”
“Is the Mesa Trail 2 waterproof and how do I clean it?”
“Compare Mesa Trail 2 vs Ridge Runner 1—what’s better for wet trails?”
“Buy Mesa Trail 2 in size 9.” (Should return a buy link with UTMs.)

If the agent waffles, add attributes to your /api/agent/attributes. If it links to the wrong PDP, tighten your product titles and collection copy.

Step 7 — Tag everything for attribution

You need to prove revenue. Use consistent, unique UTM tags for all assistant hand‑offs and mirror them in server‑side events.

utm_source=business_agent
utm_medium=organic
utm_campaign=assistant_buy
utm_content={{ sku }}

Then configure:

Landing page rules in your analytics to group all utm_source=business_agent traffic.
Checkout attribution: store UTMs in a first‑party cookie and pass them to orders (note attribute or metafield).
Server events: fire a purchase event with source=business_agent and the same UTMs.

Want a deeper implementation? See Assistant Attribution 101.

Copy‑paste snippets (Shopify‑friendly)

Add a pre‑cart link to any PDP

<a href="/cart/add?id={{ variant.id }}&quantity=1&utm_source=business_agent&utm_medium=organic&utm_campaign=assistant_buy&utm_content={{ variant.sku }}">Buy this variant</a>

Expose a tiny JSON endpoint via theme app extension (Liquid)

{% layout none %}
{
  "sku": "{{ product.selected_or_first_available_variant.sku | escape }}",
  "in_stock": {{ product.selected_or_first_available_variant.available | json }},
  "price": {{ product.selected_or_first_available_variant.price | money_without_currency | replace:",","" }},
  "currency": "{{ cart.currency.iso_code }}",
  "url": "{{ product.selected_or_first_available_variant.url | prepend: shop.url }}"
}

Note: For production, move this to an app/edge function and add caching + auth.

Compliance and trust checks

Price parity: price in schema, endpoint, and checkout must match.
Availability honesty: do not promise backorder windows you can’t meet.
Policy links: returns, warranty, shipping linked from PDP and /kb/assistant.
User consent: respect region‑specific notices for tracking and personalization.

Troubleshooting

Agent suggests OOS items: verify your availability endpoint/cache TTL and Shopify inventory policy.
Wrong variants: standardize option names (e.g., “Color”, “Size”) and avoid duplicate titles.
No buy link surfaced: ensure links are absolute, live, and under 2,000 characters total with UTMs.
Thin answers: expand /kb/assistant with concise specs and care instructions pulled from PDPs.

Roadmap after day two

Advanced ranking: introduce rules like margin‑aware recommendations and seasonal boosts.
Cross‑channel prep: the same endpoints fuel other assistants. See our overview of ambient channels in January 2026 here.
AASO: reinforce ranking signals for assistants—read our playbook AASO in 2026.

Done‑for‑you option

If you’d rather skip the plumbing, the HireNinja team can ship this implementation for you—schema, endpoints, QA queries, and attribution—within 48 hours.

Talk to HireNinja or request a demo.

Recap checklist

Define intents, guardrails, tone of voice.
Expose product, availability, and attribute endpoints; generate a buy link with UTMs.
Tighten Product + Offer JSON‑LD; keep parity with endpoints and checkout.
Publish a compact /kb/assistant page.
QA with high‑intent queries; fix gaps.
Tag and track: UTMs, server events, and order‑level notes.

Business Agent is a new front door for high‑intent shoppers. With the pieces above, you’ll greet them with precise answers, a fast path to purchase, and clean attribution back to your revenue model.

HireNinja: Blog

recent posts

about