AI Shopping & Agentic Commerce: What ChatGPT, Amazon and Google's Shopping Agents Mean for Brands

In early 2026, OpenAI relaunched “Buy it in ChatGPT,” Amazon expanded Rufus and Alexa+ into full shopping assistants, and Google began rolling agentic checkout into Gemini. For the first time, a real share of purchase journeys can start and finish inside an AI chat window — without the shopper ever landing on your website. This guide explains what's actually live today, the two competing technical standards brands need to know about, and the practical steps to make sure your products show up (and sell) when an AI agent goes shopping on a customer's behalf.

How AI shopping agents actually work

Today's AI shopping agents fall into three rough categories. Conversational discovery (ChatGPT, Gemini) lets a user describe what they want in plain language, and the assistant searches retailers, compares options, and surfaces a shortlist with prices and links. In-app shopping assistants (Amazon's Rufus, Alexa+) operate inside a single retailer's ecosystem, using the retailer's own catalog and account data to recommend and reorder products. Agentic checkout is the newest layer: rather than handing the user a link, the assistant completes the purchase itself — entering payment details, applying the user's saved preferences, and confirming the order — using a standardized protocol to talk to the merchant's commerce platform.

For marketers, the practical shift is this: the “storefront” a customer interacts with may no longer be your website at all. It may be a structured data feed your systems expose to an agent, with your brand's name, images, and price shown inside someone else's chat interface.

ACP vs UCP: the protocol fight brands need to track

Both protocols are early and evolving quickly. The realistic posture for most marketing teams in 2026 is not to pick a side, but to make sure the underlying product data both protocols depend on — accurate, structured, real-time feed data — is in good shape. That investment pays off regardless of which protocol (or both) ends up mattering for your category.

Product feed and structured data requirements

AI agents make purchase decisions based on what they can read programmatically, not what looks good on a product page. The baseline checklist:

• Implement schema.org/Product and schema.org/Offer markup on every product page, including price, currency, availability, and SKU.
• Keep your product feed (Google Merchant Center or equivalent) synced with live inventory — agents that recommend out-of-stock items damage trust fast.
• Write product titles and descriptions in plain, specific language (material, size, use case) rather than brand-voice copy that an agent can't parse for comparison.
• Ensure return policy, shipping cost, and delivery windows are machine-readable — these are now common deciding factors in agent comparisons.
• Audit for duplicate or conflicting price listings across marketplaces; agents will surface the lowest price they find, even if it's outdated.

Measuring agentic-commerce traffic and conversions

Attribution for agentic commerce is where AI referral traffic was a year ago: mostly invisible to standard analytics. A few practical steps while the ecosystem matures:

• Create unique, taggable checkout links or discount codes for any agent-protocol integrations you enable, so completed orders can be matched back even if referrer data is stripped.
• Watch for new entries in your direct-traffic and referral reports that correlate with spikes in agent-platform usage — this is the same “dark funnel” pattern covered in our AI referral traffic guide.
• Compare order values and conversion rates for any segment you can isolate as agent-originated against your overall averages — early reports of dramatically higher conversion are based on small samples and should be treated as directional, not definitive.
• Loop in finance/ops: agentic checkout may introduce new payment processors or settlement flows that need reconciliation.

Early case studies: what's actually working

Public case studies remain limited and mostly self-reported by platforms. A few patterns worth noting: retailers with already-clean Google Merchant Center feeds reported the smoothest early integrations with agentic surfaces, since the structured data work was already done. Brands selling commoditized, easily-compared products (electronics accessories, household goods) saw agents surface them more readily than brands selling highly differentiated or experiential products, where an agent's ability to “compare” breaks down. Several early adopters also reported a wave of agent-driven traffic attempting to scrape pricing and availability data far more frequently than human visitors — a reminder to make sure your infrastructure can handle increased automated load gracefully rather than blocking it outright.

What this means for your team this quarter

You don't need to pick ACP or UCP today, and you don't need a dedicated “agent commerce” initiative. What you do need is the same foundation that helps with GEO and AI search visibility generally: clean structured data, accurate and current product information, and a measurement plan that doesn't assume every visitor arrives the way they used to. Treat agentic commerce as a forcing function to fix product data hygiene that was probably overdue anyway.

Frequently asked questions