By KEY TAKEAWAYS
- Agentic commerce means AI assistants find, compare and (increasingly) complete purchases for shoppers, so your product data matters as much as your website.
- As of October 2026 the picture is messy: Google is pushing checkout inside AI Mode and Gemini via its Universal Commerce Protocol, while OpenAI pulled back from native ChatGPT checkout in favor of discovery and merchant-run apps.
- The practical move is not to bet on one protocol but to make your catalog, policies and reviews clean, structured and consistent so any agent can read and recommend you.
Picture your best customer. Now picture them never visiting your website, never seeing your ad and never reading your carefully tuned product page. They typed “best waterproof hiking boots under $200, wide fit” into an AI assistant, got three recommendations and bought one.
Was your brand one of the three? That question is what agentic commerce is about, and it’s arriving faster than most brand roadmaps assumed. It’s also messier than the headlines suggest, because the biggest players disagree on how it should work.
Here’s what’s real, what’s hype, and what to do about it.
How we wrote this: facts are linked to the companies’ own announcements or reputable reporting. Where we interpret, we say “our read.” Some of these sources were hard to verify directly at time of writing, so check live documentation before you build anything. Everything is as of October 2026.
What is agentic commerce, in plain English?
Agentic commerce is shopping where an AI agent does the legwork. It researches options, compares specs and prices, applies the shopper’s preferences, and in some setups completes the purchase. The shopper sets the goal; the software does the clicking.
That’s different from the old “AI recommends a product” idea in one way: the agent can act. Reading a product feed is passive. Holding a payment token and checking out is not.
The three layers to keep straight:
| Layer | What the agent does | What it needs from you |
|---|---|---|
| Discovery | Finds and ranks products for a request | Clean product data, strong reviews, clear positioning |
| Decision | Compares options against the shopper’s constraints | Accurate specs, pricing, shipping and returns info |
| Transaction | Completes checkout on the shopper’s behalf | Supported checkout path and payment integration |
Most of the action today is in the first two layers. The third is where the fighting is.
Where do the big platforms stand right now?
Google is pushing in-chat checkout; OpenAI has shifted toward discovery and merchant apps. The two strategies are drifting in different directions, which is why a “pick the winner” plan is risky.
Google. In January 2026 Google announced the Universal Commerce Protocol (UCP), an open standard meant to cover discovery, buying and post-purchase support. It was announced with retail and payments partners and powers a planned checkout for eligible U.S. merchants in AI Mode and the Gemini app (CNBC; Google). Google also described pilot ad formats for AI Mode and a Business Agent that lets retailers field shopper questions inside Search. Initial checkout used Google Pay, with PayPal said to be coming.
OpenAI. In September 2025 OpenAI launched Instant Checkout with Stripe and an open Agentic Commerce Protocol (ACP), starting with Etsy sellers (OpenAI). By March 2026, CNBC reported that OpenAI was revamping ChatGPT shopping after Instant Checkout struggled, with OpenAI saying the first version lacked the flexibility it wanted and that merchants could use their own checkout experiences (CNBC). Trade coverage since then describes a focus on product discovery plus retailer apps inside ChatGPT. Accounts differ on exactly how final the change is, so treat the current status as fluid.
Shopify. Shopify merchants can syndicate catalogs to AI platforms through its agentic storefront feeds, so many sellers won’t touch a protocol directly (Practical Ecommerce). If you’re on Shopify, check your admin settings for which AI channels are switched on.
Our read: the transaction layer is a land grab, and land grabs rarely end with one winner. The discovery layer is where every platform needs the same raw material: your data.
Reality check: what has shipped vs. what’s promised
Discovery is real and live. Seamless, universal in-chat checkout is not. A fair scorecard as of October 2026:
| Claim | Status | Our confidence |
|---|---|---|
| AI assistants recommend products with links | Shipped and widely used | High |
| Merchants can feed catalogs to AI platforms | Shipped (Shopify, Merchant Center, ACP feeds) | High |
| Checkout inside Google AI Mode / Gemini | Announced, rolling out to eligible U.S. merchants | Medium |
| Native checkout inside ChatGPT for everyone | Pulled back and reworked in 2026 | Medium (status changes often) |
| Agents buying autonomously without human sign-off | Mostly demos and pilots | Low |
| Most revenue shifting to agent-completed purchases | Not evidenced | Low |
We don’t have reliable data on what share of your own revenue will flow through agents, and anyone quoting a precise number is guessing. Measure your own referral traffic from AI assistants before you invest heavily.
The fear, the gain and the pain
Here’s the fear / gain / pain frame, because each is true.
Fear: disintermediation. If an agent picks your product from a shortlist, the shopper may never see your brand story, bundles or upsells. Commodity sellers feel this first. When an agent compares ten near-identical items on price and delivery, “we have a lovely About page” doesn’t help.
Gain: a new front door. Brands with clear products, honest reviews and strong signals get recommended without paying for the click. Small brands that are better documented than their giant competitors can punch above their weight. That’s the same opening we described for AI search visibility.
Pain: the plumbing is confusing. Two protocols, several platforms, shifting features and vendors who all want to sell you a “readiness audit.” You don’t need to understand every spec. You need to get your fundamentals in order, which is less glamorous and more valuable.
What does “agent-readable” actually mean?
An agent-readable brand has complete, consistent, structured product data and policies that software can parse without guessing. Agents don’t read vibes. They read fields.
Our checklist:
- Titles and attributes: specific, not poetic. “Men’s waterproof hiking boot, wide fit, size 7–14” beats “Trailblazer Pro.”
- Prices, stock and variants: accurate and updated automatically. A feed that says “in stock” for a sold-out item teaches agents not to trust you.
- Shipping, returns and warranty: stated plainly, on the page and in the feed.
- Structured data: product, offer, review and organization markup on your site.
- Reviews: recent, real and specific. They’re a major trust signal for recommendations.
- FAQ content: answer the questions buyers ask (“Does it fit wide feet?”) in plain text on the product page.
- One source of truth: the same facts in your site, Merchant Center, marketplace listings and social shops.
- Machine access: make sure legitimate AI crawlers aren’t accidentally blocked if you want to be cited. (Blocking is a legitimate choice; just make it deliberately.)
This overlaps heavily with classic e-commerce SEO, which is the good news. It’s the same work, with higher stakes for sloppiness. For the wider picture, see our AI marketing guide.
A worked example (hypothetical)
Take an imaginary home-coffee brand, “Bean There,” selling grinders. A shopper asks an assistant: “quiet burr grinder for a small apartment, under $150.”
Agent-readable Bean There: the product page states “noise level” in a spec field, the feed lists price, stock and a 30-day return policy, and 40 reviews mention “quiet” in plain language. The agent can match the constraint and justify the pick.
Agent-invisible Bean There: the product is called “The Morning Ritual,” noise isn’t mentioned anywhere and the return policy lives in a PDF. The agent can’t verify the constraint, so it recommends a competitor whose page says it clearly.
Same grinder. Different outcome. (This is an illustration, not a client result.)
How do you connect your own tools to agents?
The emerging plumbing is open protocols that let agents call your systems. Besides ACP and UCP, agents increasingly use the Model Context Protocol (MCP) to read from and act on business tools. If you build your own assistant or concierge, it’s worth understanding MCP in plain English; we cover it in MCP for Marketers. And if you want help wiring product data, support answers and order status into agents safely, that’s what our AI agents and automation work is for.
Start narrow: a product-advice assistant on your own site that uses the same clean catalog is a good proving ground, because anything it can’t answer, an external agent can’t either.
What this means for your business
Three takeaways:
- Your catalog is now marketing. Data quality used to be an operations problem. It’s now a visibility problem.
- Don’t bet on one protocol. The big platforms may standardize or fragment. Clean data works with all of them.
- Own the relationship. Even if an agent makes the first sale, your email list, loyalty program and post-purchase experience bring the second. Capture the customer, not just the order.
What to do this week
- Audit 20 product pages. For each, can a stranger (or software) find price, key specs, shipping, returns and a clear “who it’s for” statement in under 30 seconds? Fix the worst five.
- Check your feeds. Compare your Merchant Center or Shopify catalog against the live site for price, stock and title mismatches. Log every gap.
- Ask the assistants. Run ten realistic shopping prompts in ChatGPT, Gemini and Perplexity. Record whether you appear, who does and what sources are cited.
- Review your AI-crawler settings. Decide deliberately which bots you allow in robots.txt and whether your hosting or CDN blocks them by default.
- Pick one owner. Someone should be responsible for product data quality. If it’s “everyone,” it’s no one.
We keep a free automation blueprint pack with workflows for keeping feeds and listings in sync. Grab it, then spend an afternoon on your product pages. The agents are already shopping.
Frequently asked questions
What is agentic commerce?
+
Agentic commerce is shopping where an AI agent does part or all of the work for the buyer: researching options, comparing prices and specs, applying preferences, and sometimes completing checkout. Instead of a person browsing ten tabs, the agent reads product data and returns a shortlist or a purchase. For brands, that means being understood by software, not just by people.
Can ChatGPT buy things for me?
+
It depends on the merchant and the current feature set. OpenAI launched Instant Checkout in September 2025, then reportedly reworked its shopping experience in March 2026 to focus on product discovery, letting merchants use their own checkout or in-ChatGPT apps. Availability is changing quickly, so check OpenAI's current documentation before promising customers anything.
What is the Universal Commerce Protocol (UCP)?
+
UCP is an open standard Google announced in January 2026 to let AI agents handle the commerce journey, from discovery to checkout to post-purchase support. Google said it was developed with retailers and payment companies. It powers a planned checkout in Google's AI Mode and the Gemini app for eligible U.S. merchants.
How do I make my products visible to AI shopping agents?
+
Start with clean, complete product data: accurate titles, attributes, prices, stock, shipping and returns. Keep your Merchant Center or Shopify catalog in sync, mark up pages with structured data, and answer common buyer questions on the page. Agents can only recommend what they can read and trust, so consistency across channels beats clever tricks.
Will AI shopping agents replace my website?
+
Not soon, and probably not entirely. Current evidence suggests agents are strongest at discovery and comparison, while many purchases still finish on the merchant's own site or app. Your website is shifting toward being a proof hub and transaction point. Treat agents as a new front door, not a replacement for the house.
Sources
- OpenAI, Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol (September 2025)
- CNBC, OpenAI revamps shopping experience in ChatGPT after struggling with Instant Checkout offering (March 2026)
- CNBC, Google launches Universal Commerce Protocol, bets on AI-powered retail (January 2026)
- Google, New tech and tools for retailers to succeed in an agentic shopping era
- eMarketer, Google brings checkout to AI Mode as it races rivals in agentic commerce
- Practical Ecommerce, Shopify integrates AI product discovery
WRITTEN BY
Junaid Mubashar
Founder of an AI-native growth agency. Junaid plans and runs full-funnel marketing (SEO and AI search, paid social, Google Ads) and builds the AI workflows and agents that automate it. Everything on this blog is tested with real campaigns before we recommend it.
⚙️ 5 AI Automation Blueprints
Step-by-step builds for an AI lead qualifier, blog-to-social repurposer, ad report bot and more. Works in n8n, Make or Zapier.