AI Shopping Assistant Guide 2026: Agentic Commerce Protocols

AI agents have been trying to shop for a while. They scraped retailer sites, attempted to navigate checkout flows, and failed constantly. Carts broke, variants didn’t load, checkout forms changed without warning. Unlike earlier AI chatbots limited to answering questions about store policies, these AI tools can actually complete purchases. The two protocols below replace that chaos with structured APIs that give AI agents reliable, real-time access to product catalogs and checkout — automating the entire product discovery and purchase flow.

This is a format war, but not the kind where one wins and the other dies. Both will coexist. Different user bases, different discovery contexts. Brands will need to support both, the same way they run Google Ads and Meta Ads in parallel. The primary use case for each is slightly different — ChatGPT excels at conversational product discovery, while Google’s AI Mode captures high-intent search queries.

The market projections are hard to ignore:

McKinsey forecasts $900 billion to $1 trillion in US retail revenue from agentic commerce by 2030, and $3–5 trillion globally. Morgan Stanley’s AlphaWise survey shows LLM adoption already approaching 50% in the US, with AI agents capturing 10–20% of e-commerce ($190–385 billion). And Forrester predicts 20% of B2B sellers will face agent-led quote negotiations by the end of 2026.

These are not speculative numbers from AI enthusiasts. These are the research desks at McKinsey, Morgan Stanley, and Forrester.

The Basics

How a Purchase Happens

1. A user asks ChatGPT for a product recommendation using natural language (“What’s the best waterproof hiking boot under $200?”)
2. ChatGPT’s AI shopping assistant searches merchant product feeds to find relevant products
3. Products appear in conversational results with personalized recommendations and a “Buy” button
4. The user clicks Buy. Instant Checkout opens (no redirect to your site)
5. Payment processes via Stripe using a Shared Payment Token (the agent never sees the card)
6. Order is placed. The merchant fulfills and remains merchant of record

The user never leaves ChatGPT. There is no click-through to your website. The entire shopping journey and decision-making process happens inside the conversation.

Who’s Live Now

As of February 2026:

Etsy: US sellers are automatically included through Etsy’s Offsite Ads program. Live since September 2025.

Shopify: Merchant onboarding is underway. Brands like Glossier, SKIMS, Spanx, and Vuori were announced as early partners.

Instacart: Launched December 8, 2025 as the first grocery partner through the ChatGPT Apps integration.

What It Costs

OpenAI charges a 4% transaction fee on every completed Instant Checkout purchase. This was confirmed in late January 2026 when Shopify merchant onboarding began. The fee is in addition to standard payment processing (Stripe’s ~2.9% + $0.30).

For a $100 order, that’s $4.00 to OpenAI plus ~$3.20 to Stripe, about $7.20 in total platform and processing fees.

The Technical Architecture

ACP requires three components working together: a product feed, a checkout API, and a payment integration.

Product Feed: You push a gzip-compressed file (.jsonl.gz, .csv.gz, or .xml.gz) to an OpenAI-provided endpoint. Daily updates are accepted. The feed tells ChatGPT what you sell, what’s in stock, and what it costs — including complete product details like title (max 150 chars), description (max 5,000 chars), price with ISO 4217 currency code, availability status, images, and eligibility flags for search and checkout.

Checkout API: Five REST endpoints handle the purchase flow: create session, update session (shipping, variants), get state, complete purchase, and cancel. All responses return JSON with full checkout state including line items, totals, and shipping options.

Payment: If you use Stripe, enabling the Shared Payment Token is straightforward. Adyen is also referenced in the ACP spec, though Stripe is the primary integration partner. Adyen’s deeper agentic commerce role is through Google’s Agent Payments Protocol (AP2). For other payment processors, ACP offers a Delegated Payment Spec, though integration is more complex.

Platform Merchants: The Easy Path

If you’re on Shopify, you don’t build any of this. Shopify built the integration for your Shopify store. You apply at chatgpt.com/merchants, and once approved, enable it in your Shopify admin. Requirements: paid Shopify plan, Shopify Payments enabled, US-based.

If you’re on Etsy, you’re already live. US Etsy sellers are included automatically through the Offsite Ads program. No application needed.

What’s Coming in 2026

Multi-item carts (currently single-item only), PayPal integration (announced October 2025), international expansion, and additional merchant categories.

The Basics

The Coalition Behind It

UCP isn’t just Google. It launched with backing from a coalition that signals industry-wide alignment.

Co-developers who built the protocol: Google, Shopify, Etsy, Walmart, Target, and Wayfair. (Google and Shopify are the primary engineering leads; Shopify’s engineering blog describes the others as “supported by.”)

Endorsers (20+) across payments, retail, and technology: Adyen, American Express, Mastercard, PayPal, Stripe, Visa, Worldpay, Ant International, Best Buy, Carrefour, Chewy, Flipkart, Gap, Kroger, Lowe’s, Macy’s, Sephora, Shopee, The Home Depot, Ulta, Zalando, and Salesforce.

When Visa, Mastercard, Walmart, Target, and Stripe all endorse the same protocol, that’s not a press release. It’s a market signal.

How a Purchase Happens

1. A user queries Google Search in AI Mode or uses Gemini
2. The AI agent searches UCP-enabled merchant catalogs
3. Products surface with contextual recommendations
4. The user initiates purchase (Native or Embedded checkout)
5. Payment processes via Google Pay, PayPal, or other supported methods
6. Merchant fulfills. Remains merchant of record and owns customer data

The Technical Architecture

UCP is more modular than ACP. Instead of a monolithic spec, it uses independently versioned “capabilities” under the dev.ucp.* namespace. The core functionality is organized so merchants declare which capabilities they support, and agents discover them through a /.well-known/ucp profile.

Three core capabilities at launch:

• Checkout: Payment processing, cart logic, tax calculations
• Identity Linking: OAuth 2.0-based secure agent-merchant relationships
• Order Management: Post-purchase operations, order tracking, returns, and status updates

Additional capabilities include fulfillment coordination, promotional pricing, payment authorization (AP2 Mandates), and buyer consent management. Loyalty and subscriptions are on the roadmap.

Transport bindings: UCP supports four ways for data to move between agents and merchants:

• REST: Standard HTTP API calls
• MCP: Anthropic’s Model Context Protocol (JSON-RPC)
• A2A: Google’s Agent2Agent protocol
• EP: Embedded Protocol for iframe-based checkout

This multi-transport design means UCP-enabled merchants can be discovered by agents beyond Google’s own, including those built on Anthropic’s Claude or other AI platforms. It’s designed to streamline customer interactions across any AI-powered shopping assistant, not just Google’s.

Checkout Options

Native Checkout (recommended): The purchase happens within Google’s AI surface. Lower friction, higher conversion potential. Requires implementing UCP’s checkout session API.

Embedded Checkout: An iframe loads the merchant’s own checkout UI inside the AI experience. More control over the branded experience, but higher friction.

How to Get Started

Shopify merchants: Your Shopify store is automatically included. Shopify is a coalition co-developer, and their Agentic Storefronts infrastructure handles UCP for you. No separate integration needed.

Everyone else: Whether you run a custom e-commerce store or another platform, you need a Google Merchant Center account with healthy shopping feeds (no disapprovals, complete attributes), and you join the waitlist at developers.google.com/merchant/ucp. No firm public launch date has been announced yet.

When to Prioritize Which

Start with ACP if you’re on Shopify or Etsy (near-zero effort), you already use Stripe, or you want to be live now rather than waiting.

Start with UCP if you have a strong Google Merchant Center presence, your customers are heavy Google Search users, or you’re already deep in the Google Shopping ecosystem.

Most brands will need both. The user bases are different. ChatGPT shopping is conversational discovery (“What’s the best trail running shoe under $150?”). Google AI Mode is intent-based discovery (“buy waterproof hiking boots size 10”). Different moments, different AI shopping assistants, same customer. Each delivers personalized experiences through different entry points in the online shopping journey.

This isn’t VHS vs Betamax. It’s Google Ads vs Meta Ads. Both survive. Both matter. And they’re not alone—Perplexity and Microsoft Copilot are also building shopping surfaces, each with their own user base. The real question is total market share of AI-mediated commerce, not which protocol wins.

Amazon is conspicuously absent from both ACP and UCP. No endorsement. No partnership. No participation. For the company that controls 40% of US e-commerce, that silence is strategic. Brands need to understand what it means.

Amazon Is Building Its Own Path

Rufus AI is Amazon’s in-app shopping assistant. It has 300 million users, drives 60% higher conversion rates among active users, and generated an estimated $12 billion in incremental sales in 2025. Rufus recommends products, answers customer questions, and executes purchases, all within Amazon’s ecosystem.

Alexa+ (free for Prime members) adds agentic capabilities to voice commerce: it monitors prices, automatically purchases when thresholds hit, and executes multi-step tasks like restaurant booking and service scheduling.

Buy for Me goes further. It lets customers purchase from competing retailers directly within Amazon’s app. Amazon scrapes competitor sites, acts as intermediary, and completes checkout using encrypted customer data. Amazon is building an agent that shops the entire web, but only on Amazon’s terms.

Why Amazon Won’t Join Open Protocols

UCP and ACP directly threaten Amazon’s core business model. Amazon’s profit depends on controlling the discovery moment: product rankings, sponsored placements, subtle algorithmic nudges toward higher-margin items. An open protocol where AI shopping assistants compare products across every online store turns Amazon into just another merchant behind an API. That’s an existential flattening of the leverage that makes Amazon dominant.

Amazon has responded accordingly. It updated its robots.txt to block OpenAI’s crawlers, removing 600 million products from ChatGPT’s shopping results. It updated its legal terms to block AI agent behavior. It sued Perplexity for unauthorized purchases. The walled garden is getting taller.

What This Means for Your Brand

You now face a three-ecosystem world:

1. Amazon’s proprietary agents (Rufus, Buy for Me, Alexa+): massive reach, closed system
2. Google’s UCP (partnered with Walmart, Target, Shopify, Etsy, and 20+ others): open protocol, intent-based discovery
3. OpenAI’s ACP (ChatGPT Instant Checkout via Stripe): open protocol, conversational discovery

If you’re Amazon-first: Optimize for Rufus, not external agents. Focus on reviews, Q&A, and pricing competitiveness within Amazon’s walled garden. Watch for Amazon eventually offering partnership APIs. CEO Andy Jassy has signaled openness to “partnerships” with third-party agents, but on Amazon’s terms.

If you’re multi-channel: Prepare to maintain product data for three separate agent ecosystems. The feed structures differ. The checkout flows differ. The attribution models differ. This is the cost of omnichannel in 2026.

If you’re DTC: UCP and ACP may let you bypass Amazon entirely. If AI-powered shopping assistants can route high-intent customers directly to your checkout, you reduce your dependence on Amazon’s marketplace (and its commission structure). But your product data, schema markup, and checkout infrastructure need to be flawless.

The strategic bet: Walmart is betting on openness (UCP partnership). Amazon is betting on control (proprietary agents). One of them is right. Either way, your brand needs to be visible to all three ecosystems.

Here’s the uncomfortable truth about agentic commerce in early 2026: you can sell through it, but you can’t fully measure it. Your analytics stack was built for a world where customers click links, visit pages, and generate session data. AI-mediated purchases break every one of those assumptions. The customer behavior that drives the sale is invisible to your existing tools.

What You Can’t See

When a customer asks ChatGPT about your product category, the AI shopping assistant compares options, uses natural language processing to understand intent, makes a recommendation, and the customer checks out — all inside the chat. Your analytics platform sees none of it. No impression. No click. No session. No add-to-cart event. The customer interactions with the agent are completely invisible to you. The first signal you get is an order webhook.

That means you can’t answer basic questions: How many times was your product recommended? Which competitors was it compared against? What influenced the agent’s decision-making? Did the customer consider your brand and reject it?

Research suggests 70–90% of the shopping journey now happens before any trackable interaction. With agentic commerce, that figure approaches 100%.

What You Can See

It’s not a total blackout. The webhook data from ACP and UCP tells you what was sold, when, the order value, and which agent platform facilitated the purchase. Shopify’s Agentic Storefronts provide channel attribution in the admin. You can see whether an order came from ChatGPT, Microsoft Copilot, or Perplexity.

But there’s a chasm between “we know a sale happened” and “we understand why it happened and whether it was incremental.”

Why This Is Harder Than Podcasts

Podcast attribution faced a similar “dark funnel” problem. But podcasts at least gave brands promo codes and a measurement window of days to weeks. Agentic commerce happens in seconds, with one-click checkout that removes every identity-capture mechanism. The customer barely perceives the AI as an intermediary. They just think they decided to buy something.

Podcast attribution took five years to mature. Agentic commerce measurement is even harder, and the tools are earlier stage.

What to Do Now

Server-side tracking is non-negotiable. If you don’t have webhook infrastructure to capture ACP/UCP order events and route them to your analytics platform (GA4, Segment, a data warehouse), you’re flying blind. Set this up before you optimize anything else. Tools like Tealium already support ACP webhook ingestion.

Run incrementality tests. Attribution asks “what fraction of revenue do I credit to this channel?” Incrementality asks “how much additional revenue did this channel actually create?” The second question matters more. Approaches include geographic holdouts (disable agentic checkout in 5–10% of markets for 8–12 weeks), temporal holdouts (disable for random 24-hour windows), or cohort comparisons. Google lowered its incrementality testing threshold to $5,000 in late 2025, making this accessible to smaller brands.

Track discovery, not just conversion rates. Tools like Stackline’s AI Visibility product track how often your products appear in AI responses and how you compare to competitors. This is the agentic equivalent of impression share, similar to what we measure in AI search optimization. If you’re invisible to agents, conversions are a moot point.

Invest in product data quality above all else. In a channel where you can’t outspend competitors on ads, the quality of your structured data (descriptions, attributes, reviews, images) is your primary lever for agent recommendations. Merchants with 95%+ data fill rates on core attributes see dramatically higher AI agent visibility. The agents need rich product details and complete product catalogs to surface the right product for each customer query.

Plan for 18–24 months before you have high-confidence attribution. By late 2026 or early 2027, measurement frameworks will start to mature. The retailers who invested in infrastructure now will have the data to prove ROI. Those who waited will be guessing.

If You’re on Shopify (US)

Effort: Minimal. Configuration, not engineering.

For ACP (ChatGPT):

1. Confirm your Shopify store is on a paid plan with Shopify Payments enabled
2. Apply at chatgpt.com/merchants with your store URL, business contact, product categories, and estimated monthly order volume
3. Once approved, enable the ChatGPT sales channel in your Shopify admin
4. Automate webhook ingestion to track orders server-side

For UCP (Google): Nothing to do. Shopify’s Agentic Storefronts handle UCP automatically when it launches. Make sure your Google Merchant Center account is healthy: no disapproved products, no policy violations, feeds updated within 24 hours. This streamlines the entire setup for Shopify store owners.

If You’re on Etsy (US)

For ACP (ChatGPT): You’re already live. US Etsy sellers are included automatically through the Offsite Ads program. Verify by asking ChatGPT to search for products in your category.

For UCP (Google): Etsy is a coalition co-developer. Likely automatic when UCP goes live. Monitor for announcements.

For Everyone: Optimize Your Product Data

Whether you run an online store on Shopify, Etsy, or a custom platform, the quality of your product data determines how AI agents discover and recommend you. This is the most impactful work you can do right now to improve customer satisfaction and drive sales through AI shopping assistants.

Audit your product feed:

Write descriptions for AI, not SEO crawlers. AI agents use natural language processing to match natural language queries to relevant products. A title like “Men’s Trail Running Shoe – Blue – Size 10” gives an agent nothing to work with. A description like “Lightweight trail running shoe designed for technical terrain. Breathable mesh upper keeps feet cool on long runs. Aggressive lug pattern provides grip on loose rock and mud. Weighs 8.5 oz. Best for: trail races, mountain running, fast hiking.” gives the agent everything it needs to match the right product to a user’s intent.

Audit your schema markup. Run your product pages through Google’s Rich Results Test. Every product page should have valid Product, Offer, and AggregateRating schema. Prices must match displayed prices. Availability must update dynamically.

Clean up Google Merchant Center. For UCP readiness, you need zero disapproved products, no policy violations, feeds updated within 24 hours, automatic item updates enabled, and free listings turned on. Google supports 170 product attributes. The more you complete, the richer your signal to AI agents.

If You Need a Custom Integration

If you’re not on Shopify or Etsy and want ACP integration, this is a development project. You’ll need to build three things: a product feed endpoint serving data per the ACP spec (daily updates to an OpenAI-provided endpoint), a Checkout API implementing five REST endpoints, and a Stripe integration using Shared Payment Tokens. OpenAI runs conformance tests before approval covering feed validation, checkout flow testing, and payment processing. You’ll also want to automate feed updates so your product catalog stays in sync with your inventory in real-time.

For UCP, you’ll need an active Google Merchant Center account with healthy feeds, a capability profile published at /.well-known/ucp declaring your supported capabilities, and either a Native Checkout or Embedded Checkout implementation. Make sure the integration fits into your existing fulfillment and ecommerce workflows.

The build-vs-partner decision depends on your development resources, timeline flexibility, and whether you need custom requirements. If speed matters and you have a standard implementation, partnering with an agency that’s already built these integrations will save months.

ACP Product Feed: Required Fields

Every product in your ACP feed must include these fields. Missing any of them will cause feed validation failures.

Core required fields:

Additional required fields:

Optional fields (improve relevance and ranking): gtin (UPC/EAN/ISBN), mpn (manufacturer part number), material, weight, sale_price, additional_image_urls, color, size, gender, age_group, condition, product_category, shipping_price.

ACP Product Feed Example

A representative product object showing key feed fields. The actual feed format is flat JSONL (.jsonl.gz), CSV (.csv.gz), or XML (.xml.gz) with one row per variant:

{
  "item_id": "SKU-12345",
  "is_eligible_search": true,
  "is_eligible_checkout": true,
  "title": "Waterproof Hiking Boot - Men's",
  "description": "Full-grain leather hiking boot built for rugged trails. Gore-Tex waterproof membrane keeps feet dry in rain and stream crossings. Vibram Megagrip outsole provides superior traction on wet rock. EVA midsole cushions impact on long descents. Weight: 2.1 lbs per pair. Best for: day hikes, backpacking, scrambling.",
  "url": "https://example.com/products/hiking-boot-mens",
  "price": "189.00 USD",
  "sale_price": "159.00 USD",
  "availability": "in_stock",
  "image_url": "https://example.com/images/hiking-boot-main.jpg",
  "brand": "TrailMaster",
  "gtin": "012345678901",
  "product_category": "Apparel & Accessories > Shoes > Athletic Shoes > Hiking Shoes",
  "group_id": "SKU-12345",
  "listing_has_variations": true,
  "size": "9",
  "return_policy": "https://example.com/returns",
  "return_window": 60,
  "material": "Full-grain leather, Gore-Tex membrane",
  "color": "Brown",
  "store_name": "Example Outdoor Store",
  "seller_url": "https://example.com",
  "seller_privacy_policy": "https://example.com/privacy",
  "seller_tos": "https://example.com/terms",
  "target_countries": ["US"],
  "store_country": "US"
}

ACP Checkout API Endpoints

Stripe Shared Payment Token Flow

// The AI agent creates a shared payment token via Stripe
// and passes the granted token to your API.
// You attach it to your PaymentIntent:
const paymentIntent = await stripe.paymentIntents.create({
  amount: totalAmount,
  currency: 'usd',
  shared_payment_granted_token: sharedPaymentGrantedToken,
  confirm: true
});

Schema Markup: Product JSON-LD Example

Place this in the of every product page. This is what AI agents and Google’s systems use to understand your product data directly from your site:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Waterproof Hiking Boot - Men's",
  "image": [
    "https://example.com/images/hiking-boot-main.jpg",
    "https://example.com/images/hiking-boot-sole.jpg"
  ],
  "description": "Full-grain leather hiking boot built for rugged trails. Gore-Tex waterproof membrane keeps feet dry. Vibram Megagrip outsole for superior traction.",
  "sku": "SKU-12345",
  "gtin13": "0012345678901",
  "brand": {
    "@type": "Brand",
    "name": "TrailMaster"
  },
  "category": "Hiking Shoes",
  "material": "Full-grain leather, Gore-Tex membrane",
  "color": "Brown",
  "offers": {
    "@type": "Offer",
    "url": "https://example.com/products/hiking-boot-mens",
    "priceCurrency": "USD",
    "price": "159.00",
    "priceValidUntil": "2026-03-01",
    "availability": "https://schema.org/InStock",
    "itemCondition": "https://schema.org/NewCondition",
    "seller": {
      "@type": "Organization",
      "name": "Example Outdoor Store"
    },
    "shippingDetails": {
      "@type": "OfferShippingDetails",
      "shippingRate": {
        "@type": "MonetaryAmount",
        "value": "0",
        "currency": "USD"
      },
      "deliveryTime": {
        "@type": "ShippingDeliveryTime",
        "handlingTime": {
          "@type": "QuantitativeValue",
          "minValue": 1,
          "maxValue": 2,
          "unitCode": "DAY"
        },
        "transitTime": {
          "@type": "QuantitativeValue",
          "minValue": 5,
          "maxValue": 7,
          "unitCode": "DAY"
        }
      },
      "shippingDestination": {
        "@type": "DefinedRegion",
        "addressCountry": "US"
      }
    },
    "hasMerchantReturnPolicy": {
      "@type": "MerchantReturnPolicy",
      "applicableCountry": "US",
      "returnPolicyCategory": "https://schema.org/MerchantReturnFiniteReturnWindow",
      "merchantReturnDays": 60,
      "returnMethod": "https://schema.org/ReturnByMail",
      "returnFees": "https://schema.org/FreeReturn"
    }
  },
  "aggregateRating": {
    "@type": "AggregateRating",
    "ratingValue": "4.6",
    "reviewCount": "127"
  }
}
</script>

For products with size or color variants, use the ProductGroup schema type with hasVariant arrays. See the full spec at schema.org/ProductGroup.

Further Implementation References

For developers building custom integrations, these specs contain the full request/response examples and field-level documentation:

• ACP Checkout API (full request/response JSON for all 5 endpoints): developers.openai.com/commerce/specs/checkout
• ACP Product Feed Spec (complete field reference): developers.openai.com/commerce/specs/feed
• Google Merchant Center Attributes (170+ product attributes with format requirements): support.google.com/merchants
• UCP Capability Spec (checkout session API, identity linking, order management): ucp.dev

Glossary

Official Documentation

• OpenAI Commerce Developers
• ACP Product Feed Spec
• ACP Checkout Spec
• ChatGPT Merchants Portal
• Agentic Commerce Protocol
• Google UCP Developer Docs
• UCP Spec
• Shopify UCP Engineering
• Google Merchant Center Help

Announcements & Coverage

• OpenAI: Buy it in ChatGPT
• Stripe: OpenAI Instant Checkout
• PayPal-OpenAI Partnership
• Google Blog: Agentic Commerce
• TechCrunch: Google UCP Announcement
• PYMNTS: Shopify Merchants to Pay 4% Fee

Research & Analysis

• McKinsey: Agentic Commerce Forecast
• Morgan Stanley: Agentic Commerce Market Impact
• Forrester: 2026 B2B Predictions
• BCG: Agentic Commerce is Redefining Retail

Amazon & Attribution

• Amazon blocks ChatGPT shopping agent — TechRadar
• Amazon faces a dilemma: fight AI shopping agents or join them — CNBC
• Brandlight: The New Dark Funnel — How LLMs Are Hiding Your Customers’ Journey
• Tealium: Agentic Commerce and Server-Side Tracking
• IAB: Guidelines for Incremental Measurement in Commerce Media

Payment Platforms

• Mastercard Agent Pay
• Visa Intelligent Commerce
• Instacart ChatGPT Integration
• Klarna Agentic Product Protocol

What is agentic commerce?

Agentic commerce is a new ecommerce sales channel where AI agents like ChatGPT and Google’s Gemini can discover products, compare options, and complete purchases on behalf of users — all within the AI interface. Two protocols enable this: OpenAI’s ACP (Agentic Commerce Protocol) and Google’s UCP (Universal Commerce Protocol). Unlike traditional chatbots that only answer basic customer queries, these AI-powered agents handle the entire transaction from product discovery to checkout.

How does ChatGPT Instant Checkout work?

ChatGPT Instant Checkout lets users buy products directly within ChatGPT conversations. When a user asks for product recommendations, the AI shopping assistant searches merchant product feeds, displays results with a Buy button, and processes payment via Stripe — without redirecting to the merchant’s website. It uses artificial intelligence and natural language processing to understand what the customer wants and surface relevant products from its product catalog.

What is the difference between ACP and UCP?

ACP (Agentic Commerce Protocol) is OpenAI and Stripe’s protocol, live since September 2025 in ChatGPT. UCP (Universal Commerce Protocol) is Google’s coalition-backed protocol announced January 2026, coming to Google Search AI Mode and Gemini. Most brands will need to support both protocols. ACP excels at conversational product discovery, while UCP captures high-intent search queries — different use cases, but both critical for ecommerce brands.

How much does ChatGPT Instant Checkout cost retailers?

OpenAI charges a 4% transaction fee on every completed Instant Checkout purchase, in addition to standard payment processing fees (Stripe’s ~2.9% + $0.30). For a $100 order, total platform and processing fees are approximately $7.20. There are no upfront costs or monthly fees — pricing is purely transaction-based.

Is Amazon part of agentic commerce protocols?

No. Amazon has not joined ACP or UCP. Instead, Amazon is building proprietary AI shopping agents (Rufus AI, Alexa+, Buy for Me) within its own ecosystem. Brands now face a three-ecosystem world: Amazon’s proprietary agents, Google’s UCP, and OpenAI’s ACP.

How do brands get their products to show up in AI shopping assistants?

For ChatGPT (ACP), Shopify merchants can enable the ChatGPT sales channel directly from their Shopify admin — it takes minutes and requires no code. Non-Shopify merchants need to integrate with Stripe and build an ACP-compliant product feed. For Google (UCP), the protocol is still in developer preview, but brands should start by ensuring their product data is structured with complete schema markup, clean Google Merchant Center feeds, and detailed product attributes. The brands with the most complete, well-structured product data will be the ones AI agents surface first.

AI shopping assistants are a new sales channel. Not a future one, a current one. The brands that invest in structured product data, protocol readiness, and measurement infrastructure now will have a meaningful advantage as this ecommerce channel scales.

The brands that wait will be optimizing their product feeds for AI agents in 2027, watching competitors who started in 2026 collect the early data, the early customer relationships, and the early learnings.

Need help preparing your product catalog for AI agents? Whether it’s feed optimization, schema markup, ACP/UCP integration, or building a measurement framework for agentic commerce, get in touch with Opascope. We specialize in AI search optimization and GEO/AEO services for e-commerce brands, and we can help you move fast without building from scratch. See our case studies for examples of how we’ve helped brands improve visibility.

Published by Opascope | February 2026