70% Of Top Retailers Are Invisible To Agentic Commerce – Here’s Why

Ecommerce is entering a new phase – one where transactions can happen entirely off-site, making website visits optional, if not redundant. If your ecommerce and SEO teams are still playing the old game, there could be trouble ahead.

Of course, I’m talking about agentic commerce, where the entire customer journey – from discovery to checkout – can happen within AI. And boy do things move quickly!

In September 2025, OpenAI and Stripe announced their Agentic Commerce Protocol (ACP), allowing customers to not only research and decide which product to buy, but also complete single-item transactions right there in the chat with Instant Checkout.

In January, Google announced the Universal Commerce Protocol (UCP), “a new open standard for agentic commerce that works across the entire shopping journey – from discovery and buying to post-purchase support.”

And then in March, after only six months, OpenAI backed away from Instant Checkout, claiming that the initial version “did not offer the level of flexibility that we aspire to provide.” Instead, ACP is placing more focus on product discovery – driving valuable visibility for your brand and products within ChatGPT – while retailers can decide if and how they integrate their own checkout experiences.

In the same month, Google rolled out a UCP update that included a bunch of new checkout and catalog capabilities, including shopping carts.

So, if there’s one thing we’ve learned already, you’d better stay alert.

My honest prediction is that, sometime in the first half of next year, we’ll start hearing from some retailers that they achieve a better return on investment from agentic commerce than, say, paid social. The difference, and the competitive opportunity, will become clear.

So, I decided to do some research to see just how ready some of the top-performing retailers are for agentic commerce. And it appears a lot of them may be extremely unprepared.

But before we get to the research data, it’s important to understand why agentic commerce presents a very different challenge to retailers.

The End Of Traffic-First Ecommerce

We’ve known for a while that AI is gradually coopting more of the customer journey, particularly in the discovery, research and consideration stages.

The primary challenge for retailers has been getting into that all-important consideration set by ensuring their brand and products are recommended at key points in relevant AI conversations. Get that right, and hopefully, when the customer makes a purchasing decision, it’s your website they click through to.

For three decades now, website traffic has been at the heart of ecommerce. Online retailers have dedicated so much time, effort and money to refining the art of customer attraction – through SEO and content, through email and social media, through special offers and loyalty programs – anything that might lead more people to their product pages.

Yes, there’s more to ecommerce than just traffic. There are conversion rates and shopping cart optimizations, personalization and cross-selling opportunities and … well, it’s a long list. But these all rely on traffic. A potential customer has to land on your site before any of these other factors come into play.

However, when the customer doesn’t need to visit your website – when they can complete the entire transaction within an AI conversation – all previous ecommerce strategies begin to unravel.

On the same day as Google’s UCP launch announcement, Target put out its own press release describing what the new agentic shopping experience would look like.

“A guest starts a conversation in AI Mode or the Gemini app – for example, ‘I’m getting into working out and want to be both comfortable and stylish at the gym. Help me find cute and affordable floral leggings, in a light color.’ They’d then see different options to consider and purchase without having to leave the chat.”

So, how do you see that your products are included among those options?

The Mechanics Of Agentic Commerce

Agentic commerce doesn’t work like organic search. This isn’t about rankings where a lower position is still visible to customers who scroll far enough. You’re either in or out.

It also doesn’t work in the same way as AI citations or mentions. All that work your SEO team might already be doing to increase your brand’s visibility in AI won’t help you here.

Rather than crawling the customer-facing content on your product pages to extract the necessary information, both protocols instead draw data from the merchant feed or product feed and the on-page schema.

Therefore, agentic commerce isn’t a content optimization problem so much as a data plumbing one. The AI agents are merely conduits, piping raw product data one way for AI to relay to the customer and feeding transaction data back the other way once the purchase is complete.

This is important for two reasons.

Firstly, it means both protocols retain the retailer as the merchant of record. OpenAI and Google aren’t reselling your wares in the same way that, say, the Apple Store does.

Secondly, for agentic transactions to be successful, both protocols need detailed, accurate, and up-to-the-minute information that – as you’ll see – a lot of retailers don’t currently include in their feeds.

Right now, most retailers I come across treat their product or merchant feeds as a side project; a convenient way to update dynamic ad campaigns with basic product data. But the reason ad campaigns require so little information – a product name, an image, perhaps a special offer – is because they’re still about driving traffic back to the product page. The transaction still happens on the website.

If you want agentic commerce to complete that transaction on your behalf, it’s going to need a lot more information.

The 3 Schema Fields You Cannot Miss

When Google Gemini evaluates whether to recommend a product or not, it isn’t only concerned with basic transactional metadata. Google’s UCP documentation lists three other important signals:

1. priceValidUntil – confirming the pricing information is still current.
2. shippingDetails.deliveryTime – providing shipping information the customer will expect to see.
3. hasMerchantReturnPolicy.merchantReturnDays – supplying the agent with information on return windows, etc. to give the customer some buyer protection.

If any of those fields are absent or incomplete, Gemini won’t simply rank the product lower or less prominently. Your product won’t be included. Period.

Speed matters, too. AI doesn’t want to waste valuable time and CPU trying to figure things out. If an agentic agent calls your API and the response is slow, it’ll be less likely to use your feed next time.

The same goes for stale or inaccurate information that results in failed transactions, such as an item turning out to be unavailable even though the feed claimed it was in stock.

All these issues may cause the AI to view your feed as unreliable, leading it to recommend products from brands with more trustworthy feeds in future.

The Research Findings

So, now we know what’s required. How are some of the top retailers doing?

We identified 207 top-traffic product detail pages (PDPs) from 29 different retailers according to their organic traffic volume in Ahrefs. In fetching data from these pages, 52 were not readable (returning 403, 404, or status 0 errors), while 14 others turned out not to be PDPs (misidentified category or content pages).

This left us with 141 PDPs to audit – all currently winning the old-school SEO game.

We then scored each page against a 10-point UCP-readiness rubric based on Google’s UCP documentation. (The points in bold reflect the key signals mentioned above.)

1. Has Product / ProductGroup / IndividualProduct JSON-LD on the page.
2. Has GTIN identifier (gtin / gtin8 / gtin12 / gtin13 / gtin14).
3. Has MPN or distinct SKU.
4. Has brand object (name or @id).
5. Has price + priceCurrency on at least one offer.
6. Has availability (InStock / OutOfStock / etc.).
7. Has priceValidUntil.
8. Has shippingDetails.deliveryTime.
9. Has hasMerchantReturnPolicy.merchantReturnDays.
10. Has aggregateRating or review array.

The findings are stark.

1. The Basics Are Fine

Let’s start with the good news. Virtually all the audited PDPs cover the basics.

• 99% carry price and availability.
• 99% carry MPN or SKU.
• 96% carry brand.

Plus, all 141 of the PDPs appear to have fully adopted the product schema recommended by Google since 2014.

2. 70% Miss All Three Of The Most Important Attributes

However, the three key fields I mentioned earlier were only added to Schema.org between 2020 and 2021. And this is where things begin to fall apart.

• Only 18% carry priceValidUntil.
• Only 13% carry shippingDetails.deliveryTime.
• Only 11% carry merchantReturnDays.

It seems not everyone has kept their schema templates up to date.

3. 65% Don’t Include A GTIN Field

Unlike the three fields above, the lack of a Global Trade Item Number (GTIN) won’t necessarily prevent AI from using your feed, but it will handicap the agent’s ability to compare your offering with other retailers.

The GTIN is how Google’s UCP knows that Product X on your website is the same physical item as Product X on a competitor’s site. Without a GTIN, the AI agent won’t recognize your product as the same, treating it instead as a unique item. This means it can’t do things like compare pricing. If a customer asks, “Which retailer has the best price on Product X?” the answer won’t be you.

4. The Issue Is Configuration, Not Which Platform You Use

Only three of the audited brands scored an average of 8 or higher across their top PDPs:

It’s worth noting that none of the top-scoring brands would have needed to switch platforms or rebuild to make themselves agentic commerce ready. All they did was reconfigure their existing CMS to expose the additional fields.

And that’s probably the biggest takeaway here: Left unaddressed, the problem of incomplete feeds and schema could see your online store left behind by agentic commerce. But fixing the issues shouldn’t take a lot of effort and expense.

5. 15% Were Completely Invisible To Agentic Commerce

One interesting finding came about because of the data we couldn’t see.

Of the 207 URLs we audited, 32 returned an HTTP 403 Forbidden error, including PDPs from major brands like Adidas UK, UGG, Converse, and Christian Louboutin. Basically, our tools were unable to fetch the necessary data because these websites have defenses in place to prevent bots from scraping their sites.

Of course, deciding to block bots is a legitimate business choice. There are many reasons why a brand might want to protect its catalog from scrapers, such as stopping competitors from scraping real-time pricing data to undercut them on price.

But the same defenses that block a competitive audit tool will also block a Google or OpenAI agent trying to access product data for agentic commerce. The question is whether blocking agentic commerce is a conscious business decision or an unforeseen consequence.

Category SEO Won’t Help You In Agentic Commerce

When we built the audit sample, we were only interested in high-traffic product pages. Therefore, we filtered the Ahrefs results to arrive at a list of URLs most likely to be PDPs, stripping out category pages and other content.

However, before adding those filters, we noticed that category pages (PLPs) currently win the organic-search game over PDPs by a factor of roughly 10 to 1.

For example, Barbour’s top organic page is /gb/mens/jackets, attracting 20,248 monthly visits at the time of our search. But their best-performing PDP is the Ashby waxed jacket at 2,194. We saw this same pattern across most of the enterprise retailers in our sample.

And because ecommerce strategies are traditionally traffic-focused, a lot of SEO effort has been directed towards category pages over the years.

But agentic commerce doesn’t care about category pages because they carry none of the necessary schema. AI agents only read structured data at the product level. Barbour might rank well for “men’s wax jacket” in the SERPs and be completely overlooked by AI agents.

Fixing The Plumbing

All the data these agentic commerce protocols need would already exist in one or other of your systems, such as your ERP, PIM, or inventory management platform. Shipping delivery times, return windows, price validity dates, GTINs – it should all be there.

What’s missing is the plumbing to move this information cleanly into the merchant feed and on-page schema – not just once but continually.

For this to happen, three things need to change at an operational level:

1. Treat The Merchant Feed As Essential Infrastructure

Assess the completeness of your existing feeds. To start with, audit your top-selling products to check whether they carry all three UCP selection signals in schema and prioritize which to address first. Set an achievable target for how many products your team can optimize for agentic commerce each month or quarter.

2. Tighten Up Inventory Data Accuracy

If an agent calls your API and receives stale, incomplete, or inaccurate data that causes transactions to fail, the reliability signal degrades.

The safest course would be to configure your feeds to update in as close to real time as possible, with sub-hour or even sub-minute granularity.

3. Prepare For Loyalty And Dynamic Pricing

At launch, pricing in agentic commerce was mostly static. Agentic commerce could only offer the same item at the same price to everyone, even if some customers might normally be eligible for a discount or other deal when buying via the retailer’s website.

However, the March UCP update added an important new capability: Identity Linking. This allows the agentic agent to interact with a retailer’s website on behalf of the customer, “such as accessing loyalty benefits, utilizing personalized offers, managing wishlists, and executing authenticated checkouts.”

Of course, this means a little more work to implement and satisfy the key requirements, but the benefits should be obvious.

SEO Needs Supply Chain Thinking

For decades, ecommerce SEO has been about PDPs, categories, and content. Like the proverbial three-legged stool, get one of these wrong and your strategy falls over.

Agentic commerce doesn’t add another leg. I’d argue it’s another stool.

Unlike other AI conversations, agentic commerce doesn’t care which retailers have the best content. This isn’t about answering questions with insights and thought leadership. Categories don’t matter either, as the research showed. And while PDPs are still important, the protocol is only concerned with the schema.

These agentic commerce protocols want clean, accurate, and complete product data. Nothing else.

This requires a different strategy. But this strategy isn’t about optimization. Instead, it’s about acquiring the necessary component data from different systems, applying a certain amount of quality control, and then packaging it up to get this data where it needs to go, on time and in good condition.

This is supply chain strategy applied to SEO. And like any supply chain, this will require cross-functional ownership and clear deliverables.

While the SEO team is typically responsible for brand visibility, including within AI, the merchant team usually manages the product and merchant feeds. And then there’s whichever team is responsible for reconfiguring the ecommerce platform and implementing the necessary technologies to securely synchronize data between systems.

None of these teams are likely to have the authority or budget to drive the necessary changes alone, even if they’re fully aware of what needs to happen. Therefore, the CMO needs to make the case at board level, highlighting the growing connection between data infrastructure and commercial outcomes in AI to demonstrate why the completeness and performance of these feeds has suddenly become so important.

The winners in agentic commerce won’t be the loudest brands, or the best-ranked pages. They’ll be whichever retailer made their products easiest for an AI agent to buy.

More Resources:

Featured Image: Yellow duck/Shutterstock