# Is Your Product Data Ready for AI Shopping Agents?

AI shopping agents judge your product data, not your storefront design. A technical readiness guide to agentic commerce for Australian retailers.

For twenty years, an ecommerce store has been built for a human with a browser. Someone types a query, scans a results page, weighs the photos and the price, and decides. Almost everything in a modern storefront, the layout, the imagery, the persuasive copy, exists to win that human’s attention.
A growing share of that evaluation is now done by software. A shopper asks an AI assistant to find a pair of trail running shoes under $150 that can arrive by Friday, and the assistant reads catalogues, compares options, and shortlists three products before the person has looked at a single page. When the buyer never sees your storefront, none of the work you put into that storefront helps you. What the agent can read about your products is what decides whether you make the list.
That raises a concrete, answerable question for anyone running ecommerce at an Australian retailer: is your product data ready to be evaluated by a machine? The rest of this post works through it, grounded in how these systems are actually built and integrated.

What agentic commerce actually means for your catalogue
Agentic commerce is shorthand for a shopping journey where an AI agent does the discovery, comparison, and sometimes the purchase on the buyer’s behalf. Instead of browsing, the person states an intent, and the agent works across multiple retailers to satisfy it.
You do not need to track every development in this fast-moving space to prepare for it. The practical point is simpler. An agent cannot see your product the way a person does. It has no sense of a hero image or a well-designed grid. It reads whatever structured information it can pull from a feed, an API, or the markup on your pages, and it reasons over that. If a person and an agent look at the same product and the person can tell it is a waterproof size 10 in stock for $139 while the agent cannot, the agent moves on to a competitor whose data says so plainly.
That is the shift worth planning around. The quality of your product data has moved from a back-office concern to a front-of-house one.
Why product data is the bottleneck, not your storefront
Most retailers who worry about being ready for AI shopping reach first for the storefront: a chat widget, a redesign, a new theme. That is the wrong end of the problem. The storefront is the part an agent is most likely to skip.
The bottleneck sits further back, in the systems that hold the truth about each product. Your ERP knows the SKU, the cost, and the stock position. Your ecommerce platform holds the descriptions and attributes. A PIM, if you run one, enriches and distributes that content to each channel. An agent’s view of your catalogue is only as good as what those systems can express in a machine-readable form, and only as current as your integrations keep it.
This is why the retailers best placed for agentic commerce are often the ones who did the unglamorous work already: clean attributes, complete identifiers, and a real-time link between the shop and the back office. The retailers most exposed are the ones whose product data is good enough for a person to interpret, inferring from a photo or a loosely worded description, but too incomplete or unstructured for software to reason over.
The building blocks an agent relies on
Strip away the branding and the competing standards, and every approach to agentic commerce asks for the same handful of things. If these are in good shape, you are most of the way there regardless of which protocol wins.
A structured product feed. Agents start from a feed, a machine-readable export of your catalogue that lists what you sell, what it costs, and whether it is available. This is the same discipline that powers Google Shopping and marketplace listings, so many retailers already have a version of it. The bar is higher now: the feed needs to be complete, correctly typed, and refreshed often rather than assembled once and forgotten.
Product identifiers and attribute completeness. An agent comparing options leans heavily on standard identifiers, particularly GTINs, and on structured attributes like size, colour, material, and compatibility. A product with a GTIN and a full attribute set can be matched, compared, and recommended. A product described only in a paragraph of marketing prose cannot be reasoned over with any confidence. Attribute completeness is the single highest-leverage improvement most catalogues can make.
Price and stock that are accurate when it counts. An agent that recommends an out-of-stock product, or quotes yesterday’s price, produces a bad outcome for the shopper and for you. This does not mean regenerating your whole feed every few minutes, which would be wasted effort. The feed is a periodic snapshot for discovery; the figure that has to be right is the live one at the moment an agent checks availability or completes a purchase, and that is validated by a query against your source of truth rather than by how often the feed is rebuilt. Retailers with a solid ERP and inventory integration already have that link in place; retailers relying on manual syncs between systems do not.
Schema markup on your pages. Structured data in your page markup, principally Schema.org Product markup covering price, availability, and reviews, is how agents and AI search read the pages they do visit. Google’s product structured data guidelines set out the required and recommended fields. It is well-established, inexpensive to implement, and doubles as an SEO asset.
A checkout an agent can make sense of. Discovery is only half of it. If an agent shortlists your product, the purchase still has to complete, and the software helping the shopper is increasingly the thing working through your checkout to get there. You don’t need a separate checkout just for agents. What matters is that your existing checkout is something an agent can read and progress through: clear steps, meaningful labels, a predictable structure, and no dependence on a human intuiting their way past a fragile, script-heavy flow. This is the same thing that makes a checkout good for people. A clear, well-structured checkout has fewer surprises, better accessibility, and fewer abandoned carts, whether the shopper is a person or acting through an agent.
The protocol landscape, as it stands in 2026
Several standards are being built to connect agents to merchants, and the picture will keep moving through 2026 and beyond. You do not need to bet on one today. It is enough to know they exist and that they draw on the same building blocks above. Treat the detail here as a snapshot, and confirm the current state before you implement.

					Protocol
					Backers
					What it covers

					ACP (Agentic Commerce Protocol)
					OpenAI and Stripe
					Product feed, cart, and an agentic checkout, first used by Instant Checkout in ChatGPT

					UCP (Universal Commerce Protocol)
					Google with Shopify and other retailers
					Discovery, buying, and post-purchase across the shopping journey, wired into Google’s AI Mode

					AP2 (Agent Payments Protocol)
					Google with payment partners
					Cryptographically signed authorisation so an agent can pay on a person’s behalf

The important thing is that these are complementary layers, not rival products you must choose between. Discovery, checkout, and payment authorisation are different jobs, and a merchant can support more than one standard at once. All of them start from a structured product feed and accurate, current product data, which is exactly why the readiness work pays off no matter how the standards settle.
For the source material, see the Agentic Commerce Protocol specification maintained by OpenAI and Stripe, and Google’s Universal Commerce Protocol announcement.
Where each platform stands
Your starting position depends heavily on the platform you run, because each one exposes product data and feeds differently.

					Platform
					Readiness position

					Shopify
					Furthest along for out-of-the-box agent visibility, given its role in early ACP and UCP work and its native feed handling. Merchants with clean feeds can often participate quickly.

					Adobe Commerce (Magento)
					Strong native product data model and feed capability, but realising it depends on how well the catalogue and integrations were built. Flexible, and rewards good engineering.

					BigCommerce
					Solid API and feed support, with agentic readiness following its usual pattern of open, well-documented endpoints.

					WooCommerce
					Capable through plugins and custom work, but the least turnkey. Feed quality and real-time data depend on the extensions and integrations in place.

The pattern across all four is the same. The platform sets a floor, but your data quality and integrations set the ceiling. A well-built Adobe Commerce store will out-perform a neglected Shopify one, and a neglected store on any platform will struggle regardless of what the platform is technically capable of.
A product-data readiness checklist
If you want a concrete place to start, work through the following. Most retailers find the gaps are in data completeness and in the integrations that keep price and stock current, rather than anything exotic.
On the feed:

Do you produce a structured product feed today, and does it cover your full catalogue rather than a subset?
Is the feed regenerated on a sensible schedule so the discovery snapshot stays reasonably current?
Are the required fields populated and correctly typed, rather than left blank or stuffed with placeholder values?

On the data itself:

Do your products carry standard identifiers, GTINs in particular, wherever they exist?
Is your attribute coverage complete and consistent across the catalogue, or does it thin out on older or bulk-loaded products?
Is there a single source of truth for each attribute, so the same value cannot disagree between systems?

On the systems:

Can price and stock be checked against the source of truth at the point an agent checks or buys, rather than depending on a manual or periodic sync that drifts out of date?
Is Product schema markup present and valid on your product pages?
Is your checkout clear and well-structured enough for software to work through, rather than relying on a human to click their way past a fragile flow?

If you cannot answer most of these confidently, that gap is the work, and it is the same work whether or not agentic commerce grows as quickly as its proponents expect.
The same work pays off in search and AI visibility
None of this is a bet solely on AI shopping agents taking off, which matters if you are weighing the effort against an uncertain payoff.
Clean, structured, complete product data is the same foundation that makes on-site product search work well, because a search engine reasons over your attributes just as an agent does. It is also what determines whether AI search tools like ChatGPT and Google’s AI Overviews can cite and recommend your products when a shopper asks them a buying question. The feeds, the identifiers, the attributes, and the schema markup all serve discovery across every one of these channels at once.
The readiness work has three separate payoffs, and agentic commerce is only one of them. Even in the pessimistic case where agents stay a niche, you get better site search and stronger visibility in AI search for the same investment.
Where Fontis fits
Most of the noise about AI shopping is written by people selling a storefront feature or reporting on the trend. The work that actually determines whether an agent can find and recommend your products sits in the feeds, the integrations, and the data model, which is the layer we have engineered for Australian retailers for close to two decades.
We approach this as an integration problem, because that’s really what it is. Our starting point is your stack as it exists: how product data moves from your ERP through your platform and out to each channel, where it goes stale, and where the attribute and identifier gaps are. From there the fixes are concrete, whether that is a proper backend integration to keep price and stock live, or ERP-driven product data that is complete enough for software to trust.
If you want to know where your product data stands, we run a product-data and feed readiness audit that tells you exactly which gaps matter and what it takes to close them. Talk to us if that would help.

{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","acceptedAnswer":{"@type":"Answer","text":"Not in the way the headlines suggest. Most current agentic commerce models handle discovery and comparison, then send the shopper to your own site to complete the purchase. That keeps the customer relationship, the login, the checkout, and the loyalty data with you. The change is that a machine is increasingly the first thing to evaluate your catalogue, so your product data has to be good enough for software to read and trust. Your storefront still matters; it is just no longer the only thing that decides whether a shopper considers you."},"name":"Will AI shopping agents replace my ecommerce store?"},{"@type":"Question","acceptedAnswer":{"@type":"Answer","text":"No. Shopify is furthest along for out-of-the-box agent visibility because of its role in the early protocol work, but readiness is driven far more by the quality of your product data than by the platform badge. Adobe Commerce, BigCommerce, and WooCommerce can all support agentic commerce well. A well-built store on any of them, with a complete feed, standard identifiers, full attributes, and real-time price and stock, will out-perform a neglected store on a platform with better native support. What the platform gives you natively counts for far less than how well your product data and integrations are built."},"name":"Do I need to be on Shopify to be ready for AI shopping agents?"},{"@type":"Question","acceptedAnswer":{"@type":"Answer","text":"A product feed is a structured, machine-readable export of your catalogue that lists what you sell, its attributes, its price, and whether it is in stock. It is the same kind of file that powers Google Shopping and marketplace listings. AI shopping agents start from feeds like this to discover and compare products, so the feed's completeness and accuracy, and keeping it reasonably current, largely determine whether your products are eligible to be recommended. It does not need constant regeneration; it does need to be a full, correct snapshot on a sensible schedule. A feed that is incomplete or badly out of date means an agent either skips your products or works from the wrong information, which pushes the shopper to a competitor."},"name":"What is a product feed and why does it matter for AI agents?"},{"@type":"Question","acceptedAnswer":{"@type":"Answer","text":"They are complementary standards that cover different parts of an agent-driven purchase, not rival products you must choose between. ACP, the Agentic Commerce Protocol from OpenAI and Stripe, covers the product feed, cart, and checkout, and underpins Instant Checkout in ChatGPT. UCP, Google's Universal Commerce Protocol, covers discovery through to post-purchase and is wired into Google's AI Mode. AP2, the Agent Payments Protocol, handles the payment authorisation so an agent can pay on a person's behalf. All of them draw on the same foundation of a structured feed and accurate product data, so preparing that foundation readies you for any of them. This area is still moving, so confirm the current state before implementing."},"name":"What is the difference between ACP, UCP, and AP2?"},{"@type":"Question","acceptedAnswer":{"@type":"Answer","text":"Work through three questions. First, do you produce a complete, correctly structured product feed and regenerate it on a sensible schedule? Second, do your products carry standard identifiers such as GTINs and a complete, consistent set of attributes across the whole catalogue rather than just the newest lines? Third, is live price and stock validated against your source of truth at the point an agent checks or buys, rather than being reconciled manually? If you cannot answer most of these confidently, the gaps are your readiness work. The reassuring part is that closing them also improves your on-site search and your visibility in AI search, so the effort pays off even if agentic buying stays niche."},"name":"How do I know if my product data is ready for agentic commerce?"}]}

                    AI shopping agents FAQ

                Frequently asked questions

    Will AI shopping agents replace my ecommerce store?

        Not in the way the headlines suggest. Most current agentic commerce models handle discovery and comparison, then send the shopper to your own site to complete the purchase. That keeps the customer relationship, the login, the checkout, and the loyalty data with you. The change is that a machine is increasingly the first thing to evaluate your catalogue, so your product data has to be good enough for software to read and trust. Your storefront still matters; it is just no longer the only thing that decides whether a shopper considers you.

    Do I need to be on Shopify to be ready for AI shopping agents?

        No. Shopify is furthest along for out-of-the-box agent visibility because of its role in the early protocol work, but readiness is driven far more by the quality of your product data than by the platform badge. Adobe Commerce, BigCommerce, and WooCommerce can all support agentic commerce well. A well-built store on any of them, with a complete feed, standard identifiers, full attributes, and real-time price and stock, will out-perform a neglected store on a platform with better native support. What the platform gives you natively counts for far less than how well your product data and integrations are built.

    What is a product feed and why does it matter for AI agents?

        A product feed is a structured, machine-readable export of your catalogue that lists what you sell, its attributes, its price, and whether it is in stock. It is the same kind of file that powers Google Shopping and marketplace listings. AI shopping agents start from feeds like this to discover and compare products, so the feed's completeness and accuracy, and keeping it reasonably current, largely determine whether your products are eligible to be recommended. It does not need constant regeneration; it does need to be a full, correct snapshot on a sensible schedule. A feed that is incomplete or badly out of date means an agent either skips your products or works from the wrong information, which pushes the shopper to a competitor.

    What is the difference between ACP, UCP, and AP2?

        They are complementary standards that cover different parts of an agent-driven purchase, not rival products you must choose between. ACP, the Agentic Commerce Protocol from OpenAI and Stripe, covers the product feed, cart, and checkout, and underpins Instant Checkout in ChatGPT. UCP, Google's Universal Commerce Protocol, covers discovery through to post-purchase and is wired into Google's AI Mode. AP2, the Agent Payments Protocol, handles the payment authorisation so an agent can pay on a person's behalf. All of them draw on the same foundation of a structured feed and accurate product data, so preparing that foundation readies you for any of them. This area is still moving, so confirm the current state before implementing.

    How do I know if my product data is ready for agentic commerce?

        Work through three questions. First, do you produce a complete, correctly structured product feed and regenerate it on a sensible schedule? Second, do your products carry standard identifiers such as GTINs and a complete, consistent set of attributes across the whole catalogue rather than just the newest lines? Third, is live price and stock validated against your source of truth at the point an agent checks or buys, rather than being reconciled manually? If you cannot answer most of these confidently, the gaps are your readiness work. The reassuring part is that closing them also improves your on-site search and your visibility in AI search, so the effort pays off even if agentic buying stays niche.