Agentic Commerce: Is Your Online Store Ready?

The short answer: An AI shopping assistant chooses products the way a very literal buyer would. It reads your title, attributes, price, stock, delivery and return terms, and it skips anything it has to guess. So getting ready for "agentic commerce" is mostly not a tech project. It is a product-data project, and you can score your own store in an afternoon with the table below.

What has actually launched

A lot has been announced, so here is what the companies themselves say.

The Universal Commerce Protocol (UCP). Announced on 11 January 2026 as an open standard for how an AI agent talks to a store, from discovery through checkout to after-sales. Google says it was developed with Shopify, Etsy, Wayfair, Target and Walmart, and endorsed by more than 20 others. Flipkart is on that list.
Buying inside Google. UCP powers a checkout on eligible listings in AI Mode in Search and in the Gemini app. Google's Merchant Center help says this is in early access for selected merchants and available in the United States, Canada and Australia. The retailer remains the seller of record.
Shopify Agentic Storefronts. A section of the Shopify admin where a merchant manages AI channels in one place: ChatGPT, Microsoft Copilot, AI Mode in Google Search and the Gemini app. Shopify's June 2026 release added orders and sales by AI channel to the same view. Products reach these channels through Shopify Catalog, which eligible products join by default.
The Agentic plan. Shopify also lets brands that do not run a Shopify store list products in its catalog to reach the same AI channels.

Two details matter for an Indian brand. First, Shopify's help centre says a store using its ChatGPT channel must sell to customers in the United States, though the store itself can be based anywhere. So if you export to the US, this is live for you now. Second, if you sell only in India, the buy button has not reached you yet, but the recommendation has. People already ask ChatGPT and Gemini what to buy, and those answers are built from the same product data.

The useful way to think about it: an agent is a customer who reads every field, believes exactly what is written and never looks at the pictures for clues.

The product-data scorecard

Score one product page. Give 0 if the item is missing, 1 if it is partly there, 2 if it is complete and accurate. Ten items, 20 points.

#	What the agent looks for	0	2
1	Literal title	"The Monsoon Edit"	Says what it is, the material and the size
2	Specific category	"Home"	"Double bed sheet sets"
3	Description that answers questions	Mood and adjectives only	Who it is for, what it is made of, dimensions, care, what is in the box
4	Attributes in fields, not in prose	Size and material buried in a paragraph	Size, colour, material, weight in their own fields
5	Variants grouped properly	Each colour is a separate product	One product, with colour and size as options
6	Identifiers	None	Brand, SKU and barcode (GTIN) where you have one
7	Live price and stock	Stock updated by hand	Price and availability update automatically and match the page
8	Delivery details	"Ships fast"	Cost, regions served and delivery days
9	Return policy	A line in the footer	Its own page: the window, the conditions, who pays
10	Reviews	None on the page	Rating and review count shown and marked up

16 to 20: you are ready. Keep it maintained. 10 to 15: fix the gaps on your top twenty products first. Under 10: an agent will usually find a competitor easier to recommend.

A worked example

A made-up store, Kaveri Home, and one made-up product page.

#	Item	What we found	Score
1	Title	"Indigo Dreams"	0
2	Category	"Bedding"	1
3	Description	Three lines about slow mornings	0
4	Attributes	Thread count in the text, no size field	1
5	Variants	King and queen are separate products	0
6	Identifiers	SKU only	1
7	Price and stock	Synced automatically	2
8	Delivery	"Free shipping over ₹999", no timeline	1
9	Returns	Clear policy page, 7 days	2
10	Reviews	40 reviews shown, not marked up	1
	Total		9 / 20

Kaveri Home has a good product and a store that works for humans. For an agent answering "a cotton king-size bedsheet under ₹2,500 that can be delivered this week", it is nearly invisible: nothing in a field says cotton, king or how long delivery takes.

Before and after: one description

Before

Indigo Dreams. Wake up to slow mornings wrapped in the timeless charm of hand-crafted comfort. A piece of heritage for your modern home.

After

Kaveri Home Cotton King Bedsheet Set, Indigo Block Print. A 100% cotton, 210 thread count bedsheet for king beds, hand block printed in Jaipur.

In the box: 1 flat sheet (274 × 274 cm) and 2 pillow covers (46 × 69 cm)
Fits mattresses up to 198 × 183 cm
Feel: soft and breathable, suited to warm weather
Care: machine wash cold. Colour may release slightly in the first wash
Delivery: 3 to 5 days across India. Returns: 7 days, unused

The first version answers no question a buyer would ask. The second answers eight. Keep one line of brand voice if you like, but put the facts first. Shopify's own guidance says the same thing: use literal language and keep marketing copy out of product fields.

Structured data: what each type tells a machine

Structured data is a hidden label on the page, written in a format called schema.org, that states the facts without the machine having to interpret your layout. These are the ones that matter for a store, based on Google's merchant listing documentation.

Type	What it tells a machine
Product	This page is one product: its name, images, description, brand, SKU, GTIN, colour, material, size
Offer	It is for sale at this price, in this currency, and it is in stock or not
AggregateRating and Review	How buyers rated it and how many of them did
OfferShippingDetails	What delivery costs, where you deliver and how long it takes
MerchantReturnPolicy	The return window, the conditions and the cost
Organization	Who the seller is. Google recommends putting store-wide shipping and return policies here instead of repeating them on every product
BreadcrumbList	Where the product sits in your catalogue

Google only requires name, image, price and currency. Everything else is "recommended", which is exactly why it separates stores. Test any page with Google's Rich Results Test. The same markup supports your SEO, so none of this work is wasted if agents take longer to arrive than promised.

Four ways stores get this wrong

Writing for mood, not for questions. Poetic names and descriptions work on Instagram. In a product field they hide the product.

Putting facts inside images. A size chart saved as a picture is unreadable to an agent. Put it in text or a table.

Letting data drift. If your feed says in stock and the page says sold out, the agent has been misled once. Google's guidance is blunt: keep product data as complete and current as possible.

Buying a plugin and stopping. An app can output the markup. It cannot invent the dimensions you never entered.

A 30-day plan

Week 1: Export your catalogue to a spreadsheet. Score your top twenty products. Count the empty cells.
Week 2: Rewrite those twenty titles and descriptions in the "after" format. Move sizes, materials and weights into fields. Merge split variants.
Week 3: Add or repair structured data for Product, Offer, shipping and returns. Publish returns and shipping as their own plain pages.
Week 4: Test. Run pages through the Rich Results Test, then ask ChatGPT, Gemini and Perplexity the questions your customers ask and note whether you appear and what they say about you. If you are on Shopify and sell to the US, review the Agentic Storefronts settings in your admin.

Where this fits

You rarely need a new site for this. It is a data and content job on top of a fast, well-built store, and it is part of how we approach website development on Shopify, WordPress and React. Being named in AI answers is the work of GEO and AEO, and both start from the same clean product data.