In The GEO Playbook, we explored the three pillars of AI visibility: SEO, brand, and product data. In this piece, we’re doubling down on that third pillar for technical leaders—there’s a lot to unpack, and we’re sure you have questions.
The number of conversations we’re having with enterprise decision makers about this has exploded over the past year, and for good reason.
AI is changing the commerce game, fast. AI-driven orders grew 15x year-over-year in 2025. Gartner®, in their new report “Optimize Product Data for Agentic Commerce”, also predicts that by 2030, 20% of transactions will be executed through AI platforms using on-platform check-out or by AI agents.
However, when we’re having these conversations, only a small fraction of companies we speak with have a formal strategy for handling their product data—a gap that boards and CEOs are starting to notice.
That means a question is coming (if you haven’t gotten it already): "What steps are we taking to be AI-ready?"
This piece will arm you with an answer—a concrete technical framework for understanding your options for creating AI-ready product data.
Why should you listen to us? Fair question. Here are just a few reasons:
- We co-developed UCP with Google. The Universal Commerce Protocol is an open standard for agentic commerce that benefits the entire industry, not only Shopify customers. It’s designed to tackle real commerce complexity—things like structured product data, checkout infrastructure, and fulfillment logic—which Shopify knows better than anyone.
- Leading AI platforms choose Shopify first. Brands on Shopify are among the first to sell on AI channels like ChatGPT, Microsoft Copilot, Google AI Mode, and Gemini.
- We've processed over $1.4 trillion in global commerce since our inception. That scale gives us signal. We see what's changing before most of the market does, and we're actively building for that future.
This piece reflects what we're learning, and building, in real-time.
The realities of product data at most enterprise companies
If your first reaction to “product data quality is the prerequisite for the AI era” was a quiet groan, we get it, and you’re not alone.
We talk to CTOs, CIOs, and other enterprise technical leaders every day. We hear the stories. And the reality is this: product data at most enterprise companies is a huge pain point.
Here's what we typically see.
Most enterprise retailers have evolved their data architecture over years, often decades. What started as an ERP eventually required a PIM. The PIM needed to feed the website. The website needed to connect to marketplaces. Each channel introduced its own system—OMS for orders, WMS for fulfillment, DAM for assets—and each system became a "source of truth" for some subset of product attributes.
The result? Your product data doesn't live in one place. It lives in six. Or twelve. And because every channel you sell on—your website, marketplaces, social, etc.—requires its own format of product data, you end up with multiple versions of that data in your backend systems. All of this is stitched together by middleware that, in many cases, predates half the people currently responsible for maintaining it. All this is compounded by scale—the higher your SKU count, the harder this becomes.
This isn't a failure of execution. A lot of this comes down to entropy. Each system was added to solve a real problem. Each customization made sense at the time. But over time, it inevitably compounds into technical debt that nobody fully understands.
We're not painting this picture to pile on. We're painting it because if we're going to talk about what AI-readiness actually requires, we have to start with an honest acknowledgement of what stops most CTOs before they ever even get started.
So first, let’s dive into what it actually means to have “agentic ready” product data.
What does “agentic ready” product data actually look like?
Making your product data agentic ready is a prerequisite if you want customers to discover and purchase from you in AI chats, but what does it actually take to be “agentic ready”?
In short, LLMs need data that’s machine-parsable and real-time. Let’s break that down.
Machine parsable data
Machine-parsable data is product information structured in a way that software systems—particularly AI agents and LLMs—can directly query, interpret, and act upon without human intervention or intermediary translation.
Most product data was built for humans browsing websites: readable prose, visually merchandised pages, elegant theme-level customizations that display beautifully on a storefront. But there's a gap between what renders well for a human shopper and what an AI agent can actually consume.
It turns out this isn't about data quality in the traditional sense—it's about data accessibility. A product page might have every piece of information a buyer needs, but if that information lives in Liquid templates, JavaScript rendering logic, or custom display rules, it's invisible to AI.
Real time data
Real-time data, in the context of AI commerce, refers to product information—particularly pricing and inventory—that is accurate at the moment of the shopper’s query, not at the moment the LLM last scraped your website.
When an LLM doesn’t have direct access to your product data via an API, it relies on web scraping. That means by the time a potential customer sees your data, there’s a chance that inventory has come and gone, new colors have been introduced, or new discounts have dropped. Your product information changes too frequently to rely on this method.
Now that we’ve covered what agentic ready product data actually looks like, let’s dive into what you can do about it, and the pros and cons of each path you could take.
Path 1: Optimize your product data for AI agents
We won’t bury the lede here, and this should come as no surprise, but if you want to achieve agentic-ready product data, our recommendation is to migrate to Shopify or sign up for the Agentic Plan (more on that below).
With the Agentic Plan, you get access to Shopify Catalog which automatically cleans, verifies, and syndicates your product data out to AI platforms, so that agents can find your products and make recommendations. And, with Agentic Storefronts, customers can make purchases from those recommendations.
But let's say you decide to participate actively in AI commerce and recognize the need for clean, real-time product data, and you want to do it yourself rather than migrating to Shopify or adopting our Agentic Plan.
Without a platform that takes on that technical complexity for you, here's what you're facing:
For one, each AI platform has different requirements. According to Gartner:
- OpenAI/ChatGPT: Apply via Merchant application, build product feed to their specifications
- Google/Gemini: Submit enriched feed to Google Merchant Center, augment with "GenAI Attributes," host a UCP Manifest file, build a real-time inventory check endpoint
- Perplexity: Integrate via Feedonomics, Sonar API, or their Enterprise Resource Hub
That's three separate integration projects. Three different data formats. Three different approval processes. Three different ongoing maintenance burdens. And that's just today's platforms—new AI shopping experiences are emerging constantly.
The technical debt compounds. Every hour your team spends building and maintaining AI platform integrations is an hour not spent on innovation. Every sync job that needs monitoring is attention diverted from building what actually differentiates your business.
Regardless of your commerce platform, the work starts in the same place: making your product data machine-parsable, real-time, and complete.
Here's what that actually looks like. But understand that Shopify Catalog takes away the lift of getting all of the below right:
Audit where your data actually lives: If your product information is structured or merchandised through display logic—color swatches rendered via JavaScript, product groupings managed in templates, custom options modelled as tags—you’re keeping it from being read by AI. Direct API access via Shopify Catalog is going to be the absolute best solution here, so that the most up-to-date and accurate information is directly accessible by agents in real time.
Get your variants right: If you’re using separate product pages for each variant of a product, an agent will likely assume these are all different products. That means if a potential customer asks “Does this come in red?”, they might get the wrong response back. Group genuine variants under a single parent so an agent understands that it's one product with options. That's what it needs to make high-quality recommendations.
Assign specific product taxonomy: Use the most specific product type or category possible: "men's insulated winter boots," not "footwear." The more precise your taxonomy, the more accurately AI can classify and surface your products. You need to take out the guesswork.
Use literal language: Agents are doing their best with the data provided and will often interpret literally, so keep marketing copy out of product fields. If your product title just says “Ocean breeze” instead of "Texturizing sea salt spray", an agent might not get enough context to surface your product when a customer asks for recommendations. Creative copy is still essential for human browsers, but you need to make sure you’re providing the correct level of detail for LLMs now as well.
Solve for real-time accuracy: Pricing and inventory need to be accurate at the moment of the shopper's query, not the last time a crawler hit your site. Building and maintaining real-time API endpoints for each AI platform is one of the most resource-intensive problems in agentic commerce. However, Shopify Catalog can serve real-time, validated data directly to AI platforms. When an AI agent asks "is this available at this price?", Shopify can answer instantly and accurately, which is exactly what’s required.
That's the work. And for most enterprise brands (especially those on fragmented stacks with data living across a PIM, ERP, OMS, and DAM) it's a significant lift. You're looking at auditing years of accumulated data architecture, restructuring how variants and groupings are modelled, and building or buying real-time API infrastructure that each AI platform can tap into.
This is where your platform choice matters.
On Shopify, a lot of this is handled for you. Shopify Catalog automatically syndicates your product data to AI platforms via direct API (no bespoke integrations needed). Agentic Storefronts, which is powered by Catalog, is automatically enabled for eligible merchants and connects you to ChatGPT, Microsoft Copilot, Google AI Mode, and Gemini out of the box. Allowing for your products to not just be recommended and discovered on these platforms, but actually purchased, too.
📍 A Critical Distinction: Discovery vs. Native Selling
Your products can show up in AI recommendations regardless of what actions you have or haven’t taken. That's discovery—it happens through scraping, feeds, or Shopify Catalog. But native selling—the ability for consumers to actually purchase—requires Agentic Storefronts to be enabled.
- Discovery (via scraping, feeds, or Shopify Catalog) — happens regardless of Agentic Storefronts settings.
- Native selling (checkout from a chat) — requires Agentic Storefronts to be enabled.
For brands that have custom data setups, Shopify also provides tools like the Combined Listings App to restructure your variant groupings, and Catalog Mapping to map custom fields to the standard fields that AI channels understand, so you don't have to rearchitect everything from scratch.
And if you're not on Shopify and don't want to migrate, the Agentic Plan gives you access to all of this without replacing your existing platform. Connect your product feed, configure your checkout, and you're selling across AI channels without needing a Shopify online store.
The point is this: the optimization work is the same no matter where you are. The question is how much of it you want to do yourself.
Path 2: Do nothing / wait
You might be considering this option if AI seems too nascent or risky.
The pushback we often hear: “Our product data might not be the cleanest, but it technically works. So why bother?”
But the risk isn't acting on AI readiness. The risk is not being prepared when it matters.
Think of it like this: you're making sure your house is in order today, so that you're not scrambling as a brand new sales channel takes off.
Here's what we're seeing:
- AI-driven traffic to Shopify sites grew 8x year-over-year in 2025
- AI-driven orders from search and recommendations grew 15x
- Gartner predicts that by 2030, 20% of transactions will be executed through AI platforms
The curve is early. You have time. But that time is exactly why acting now is the safe play—not the risky one.
Path 3: Rely on SEO and brand alone
Maybe you’re bought into the idea of agentic commerce, but it seems like more of a marketing problem than an engineering one.
After all, SEO and brand have huge roles to play in how you show up when people use AI agents to shop. Maybe you’re even seeing traffic already coming to your site via AI channels without any additional investment on your part.
While true, that’s only part of the equation.
Without clean product data flowing directly to these LLMs, AI platforms might still find you through web scraping—they just might not find the most accurate information about you. Your product information changes too frequently to rely on web scraping alone. Things like pricing, inventory, and product descriptions all need to be up-to-date and accurate when you're trying to make a sale.
Leaving agentic readiness to SEO and brand also means that your products might technically be discoverable in AI chats, but relying on SEO and brand alone means that shopping stops at discovery—purchases can only be completed if you navigate away from the conversation. And in modern commerce where every friction point costs you sales, this is a path we wouldn’t recommend taking.
Doing nothing, or delegating agentic commerce readiness to SEO and brand, might initially seem like a neutral choice. In reality, it’s a choice to potentially introduce unnecessary friction where you can’t afford it.
The window is now
Product data has always mattered. But for most of retail's digital history, "good enough" was actually good enough. Your website worked. Orders came through.
But that's changing.
These aren't hypothetical requirements for some distant future. They're the technical specs for commerce channels that exist today and are on the rise.
The good news: you don't have to solve this alone. Shopify takes the lift out of getting your data agentic-ready and selling on AI channels.
Want to take a closer look at AI discovery? Check out our new webinar: A practical guide to AI discovery.
