Playbook for Brands and Merchants
A practical guide for DTC and wholesale brands preparing for AI shopping agents: product data, agent surfaces, protocols, pricing, and returns.
In short: AI shopping agents choose products by reading structured data, not by looking at your homepage. For most brands the highest-return work is making product data complete and accurate, turning on the agentic channels your commerce platform already offers, and deciding in advance how you want price, brand presentation, and returns to work when an agent sits between you and the customer.
What changes for brands
In a conventional funnel, a shopper lands on your site, browses, and checks out under your brand. In agentic commerce, a shopping agent such as ChatGPT, Gemini, Copilot, or Perplexity does the browsing. It interprets a request (“waterproof trail runners under $150, wide fit”), queries catalogs, compares offers, and either hands the shopper to your checkout or completes the purchase through a protocol.
Three practical consequences follow:
- Data replaces merchandising as the first impression. The agent’s shortlist is built from attributes, prices, availability, reviews, and policies. Missing or ambiguous fields mean you are not considered.
- The checkout may not be yours. Depending on the surface, checkout happens on your site in an in-app browser, inside a merchant app in the assistant, or natively in the assistant via a protocol such as UCP or ACP. You usually remain merchant of record, but you control less of the page.
- Comparison is instant. Agents can compare your price and terms against every other seller of the same item in one step, including your own wholesale partners.
The current landscape (as of October 2026)
| Surface | How brands get in | Checkout model |
|---|---|---|
| ChatGPT | Product feeds via ACP; Shopify merchants via Agentic Storefronts | Hand-off to the merchant’s site or to retailer apps inside ChatGPT; OpenAI ended native Instant Checkout in March 2026 |
| Google AI Mode and Gemini | Merchant Center feeds; UCP integration for native checkout (US, early access) | Native checkout on eligible listings, or embedded checkout for approved merchants |
| Microsoft Copilot | Shopify Catalog, PayPal Store Sync, and other feeds | Copilot Checkout for eligible merchants |
| Perplexity | PayPal Store Sync and direct programs | Varies by merchant and payment partner |
| Meta (Muse agent) | Shopify Agentic Storefronts Meta channel | Direct checkout, on by default for eligible US Shopify merchants since September 21, 2026 |
Shopify brought Agentic Storefronts live in ChatGPT on March 24, 2026, making merchants’ products discoverable there by default; the same admin channel manages Copilot, AI Mode, and Gemini. On September 21, 2026 Shopify turned on agent direct checkout by default for eligible US merchants in its Meta channel (merchants can switch it off in the admin), and on September 28 it added WebMCP checkout tools so browser-based agents can complete checkout on eligible storefronts after the buyer authorizes it. Defaults matter: check which agent channels are already live for your store. See commerce platforms for how other platforms compare.
Opportunities
- New demand at the moment of intent. Conversational queries are long and specific. Brands with rich attributes (materials, fit, compatibility, certifications) can win queries that keyword search never surfaced.
- Lower integration cost than past channels. Most brands enable agent channels through their platform, PSP, or feed provider rather than building per-assistant integrations.
- Better long-tail visibility for wholesale brands. Product databases and catalog APIs that aggregate across retailers can surface your products wherever they are stocked. Channel3 is one example of a cross-merchant product API that agents use for search; Shopify Catalog, Google Merchant Center, and feed tools such as Feedonomics play related roles.
Risks
- Disintermediation. The assistant owns the conversation and may own the post-purchase relationship unless you secure account linking, email capture, and order notifications.
- Inaccurate representation. Stale prices, wrong variants, or hallucinated claims can appear in answers. Agents tend to trust the most structured source available, which may be a reseller rather than you.
- Price and channel conflict. Wholesale partners, marketplaces, and gray-market sellers become directly comparable to your DTC price.
- Unwanted automation. Some agents browse your site directly. Bot controls that block all automation can also block legitimate agents; see Web Bot Auth for signed-agent approaches.
Product data readiness
Product data is the foundation. Prioritize:
- Identifiers. GTIN/UPC, MPN, and brand on every variant, so agents can match the same item across sellers.
- Variant structure. Clean parent/child relationships with explicit size, color, and material values rather than values buried in titles.
- Attributes that answer questions. Dimensions, compatibility, care instructions, ingredients, certifications, and use cases written as structured fields, not marketing copy.
- Real-time price and availability. Agents increasingly check live inventory. Feed latency of a day can mean lost or cancelled orders.
- Policies as data. Shipping times, return windows, and warranty terms in machine-readable form; Merchant Center and the protocol specs both have places for these.
- Images and alt text. Multiple clean images per variant support visual and multimodal search.
Protocol enablement through platforms
Most brands should not implement protocols from scratch. A practical order of operations:
- Shopify merchants: review Agentic Storefronts settings in the admin (including the Meta channel’s direct checkout toggle), check which products are syndicated, and confirm ChatGPT referral attribution is visible.
- Other platforms: check your platform’s or PSP’s agentic programs. PayPal’s Store Sync, Stripe’s agentic commerce integrations (supporting ACP and UCP), and Adyen Agentic (announced June 2026, limited availability for US enterprises) all aim to expose catalog and checkout to agents without custom builds.
- Google: keep Merchant Center feeds healthy and join the UCP waitlist if you want native checkout in AI Mode and Gemini.
- Custom stacks: consider publishing a UCP profile at
/.well-known/ucpand an ACP-compatible feed; see protocols overview.
Brand control
Decide what “on brand” means when you do not control the page. Useful levers include: authoritative product descriptions and claims in your feed, brand-approved images, clear positioning attributes (for example “made in USA”, “vegan”), and a merchant app or embedded checkout on surfaces that allow one. Monitor how assistants describe your products by running a fixed set of prompts monthly; this is the core of agent optimization.
Pricing and MAP
Minimum advertised price (MAP) policies were written for ads and listings. Agents that compare offers across sellers make violations more visible and more consequential. Steps to take:
- Audit how your products are priced across authorized resellers in agent answers.
- Align feed prices with promotions; ACP and UCP both support promotions and discounts, so publish them explicitly rather than relying on on-site banners.
- Review reseller agreements to cover agent channels and data syndication.
- Avoid dynamic pricing that changes between the agent’s quote and checkout; protocols such as AP2 are designed around a signed cart, and a mismatch causes failures and disputes.
Returns and post-purchase
Agent-initiated orders still produce returns, and the shopper may not remember which site they bought from. UCP defines an order capability for fulfillment tracking, returns, and refunds, and Shopify exposes order webhooks to agents. Make sure return policies are machine-readable, order confirmation emails clearly name your brand, and customer service can identify orders by channel. See consumer protection and liability.
Prioritized checklist
Now
- Audit product data completeness: identifiers, variants, key attributes, policies.
- Enable agent channels available through your platform and PSP.
- Tag agent-originated orders and referrals in analytics.
- Review bot management rules so verified agents are not blocked.
Next
- Publish promotions and return policies in structured form.
- Run monthly prompt audits across major assistants.
- Update reseller and MAP agreements to cover agent channels.
- Pilot native or embedded checkout on one surface.
Later
- Implement account linking and loyalty in agent checkouts.
- Evaluate building a branded agent or merchant app inside assistants.
- Integrate agent order data into returns and customer service workflows.
Common mistakes
- Treating agent visibility as an SEO copywriting exercise instead of a data-quality project.
- Blocking all automated traffic, including signed agents from major assistants.
- Publishing feeds that update daily while prices change hourly.
- Ignoring wholesale partners’ listings, which agents may treat as more authoritative than yours.
- Assuming the assistant will pass customer data back; confirm what each surface shares.
Sources
- Agentic commerce momentum — Shopify
- OpenAI shifts checkout plans in its agentic commerce strategy — Digital Commerce 360
- Getting started with Universal Commerce Protocol on Google — Google for Developers
- Universal Commerce Protocol — ucp.dev
- Agentic Commerce Protocol — agenticcommerce.dev
- Agentic commerce — Stripe Docs
- PayPal Launches Agentic Commerce Services to Power AI-Driven Shopping — PayPal
- Adyen Announces Adyen Agentic — Adyen
- Build commerce agents with UCP — Shopify Dev
- Shopify turns on agent checkout by default for a million merchants — Forkast
- Shopify opens checkout to browser-based AI agents — TechCrunch
- Channel3 — Channel3