Hey everyone, Scott Austin here.
Today we’re talking about a major shift in e-commerce and that is AI becoming a sales channel.
And here’s the key takeaway upfront:
If your product data isn’t structured, your products won’t show up in AI-driven shopping.
This isn’t something coming someday. Shopify is already preparing for it with Agentic storefronts and the Standard Product Taxonomy. And that means the work you do now directly impacts whether your products get recommended—or ignored.
Now, a quick note before we dive in. A lot of what we’re seeing is still evolving. We don’t have years of data yet. But we do have enough signals to understand how this is going to work—and more importantly, what you should be doing right now.
So let’s dig in.
The Addition of Agentic Commerce
I’m going to use the term “agentic” instead of AI.
Agentic commerce means AI systems choose products for customers instead of customers browsing stores.
Instead of scrolling through 10 blue links on Google search results pages, a customer can just ask: “What are the best running shoes under $150 in size 13 wide?” And the AI agent gives the customer a short list of products that it determines meet the needs. Not a list of stores, but a list of products across multiple stores.
That’s a completely different game. Now, where is this actually happening today? You’re seeing it in:
- Google AI Overviews
- Chat-based shopping tools like ChatGPT and Perplexity
- Product recommendations inside search and shopping experiences
- Feed-driven platforms like Google Shopping
Now here’s the important part—these systems are not looking at your site the way a human does. They rely on three things:
- Structured product data (your taxonomy and attributes)
- Product feeds (like Google Merchant Center or the Shopify Catalog API)
- Clean, crawlable product pages
This can be a significant shift in the buyer journey. The AI agent is replacing much of the evaluation and decision making that is currently done by the customer. And the agents will make their decision differently from the way customers decide.
Now here’s the really interesting part. You may have heard about Agentic storefronts from Shopify or others. In fact, if you go to your Shopify admin and click on Settings > Sales channels, you’ll probably see Agentic storefronts listed at the bottom and ChatGPT shown as coming soon. These agentic storefronts are not like the ones we are used to with Facebook, Instagram and TikTok where our brand has its own presence within the social platform. Its more like, Google Merchant, where the store sends a data feed to Google and then Google builds out the customer’s shopping experience and our product data is just one of many brands in that experience.
Why Structure Matters
So Agents are going to have tons of products to match to customers queries. How will your products be able to stand out? Well, the answer is for your products to match to the query better than other products do and that will happen by you having better structured data for your products.
Let me illustrate with an example.
Say a customer is shopping for shoes. The promise of agents is that the customer can give them a very specific query with lot’s of context. Here’s an example input.
I’m a 59 year old man that has recently started weight lifting. My feet are size 13 and kind of wide. So I usually wear a size 13W or sometimes a size 14 shoe. I’ve got bad ankles, so I’m worried about stability. And my favorite color is purple. A dark or royal purple not a light or lavender purple. Find me a pair of sneakers under $80 that I can wear to the gym.
This is a ton of requirements from the customer for the agent to work with. And this is the power of agents, it can take all of those customer requirements written in whatever way the customer gives them. The agent can then convert the requirements into a more structured format like Color – Dark Purple. And then the agent can find the set of products that meet them all. So for your products to be considered, it will need all the data attributes mentioned or at least more data attributes than your competitors have.
To make it easier for agents to do their evaluation, the data needs to be structured, so the agent can do apples to apples comparisons. Let’s think about the structuring of the data on two levels.
· The first level is the buckets we put the data in. Think of product categories as a bucket.
· The second level is the content choices for that bucket.
And to keep the data for the agents apples to apples, both of these levels need to be standardized. In other words, all products need to use categories and choose from the same list of categories.
The Shopify Standard Product Taxonomy
And this is where the Shopify Standard Product Taxonomy or SPT comes in. The SPT has two levels of data.
· The first is the product categories which Shopify added in 2023 and continuously evolves. You can see product categories in the Shopify admin today in the product pages. For our sneaker example, the product category from the SPT that I would choose is Apparel & Accessories > Shoes > Athletic Shoes. Interestingly, there’s a sibling node to Athletic Shoes called Sneakers, so sometimes you have to apply some judgement when picking categories.
· The second level in the SPT is attributes. We see them on the product page in the Shopify admin as Category metafields. The SPT provides a list of attributes and their appropriate choices for each of the 10,000+ categories. The Athletic Shoes category has 14 defined attributes including things like Shoe size, Color and Activity. And each of the attributes comes with a defined list of choices. For example, there are 65 options under Activity.
You can look at all of the data in the SPT through a tool provided by Shopify. There’s a link to it in the show notes.
In the very near future, your product data is going to be sent by Shopify to the agents. Now the data sent is more comprehensive than categories and attributes, but brands should focus on those two in 2026 as they are the most structured data due to the SPT. Therefore, categories and attributes are the most effective way to evaluate products and give customers relevant product recommendations.
Your Product Taxonomy
So to be competitive in agentic commerce you’re going to need to set the product category for each of your products. And you’ll need to also set the category attribute values.
Shopify allows you to enhance the category attributes and the attribute values with your own additions. But I’ve found that the current agentic feed is only sending the values set by Shopify in the SPT. I expect that will quickly evolve just like we’ve seen Shopify evolve metafields quickly over the past few years. So I’m recommending that stores start with getting categories and attributes completely filled out in their catalog using just the SPT for round 1.
So let’s talk about how stores can get this data enrichment done.
Product Taxonomy in New Catalogs
For our first scenario, let’s cover a new brand that’s building out their product catalog. Shopify has some AI built into the Shopify admin called Magic. Magic will help you select the category and attributes for your products. It looks at the current data in your product like title and description to determine its recommendations. So if your product data is light, the recommendations will be low quality. You can accept the recommendations in the admin. But its done product by product, not in bulk which can be quite the manual effort especially in a larger catalog.
To make this process much easier and faster, we’ve got a Shopify app called Datify. It’s linked to in the show notes. Datify is currently in Beta and is focused on giving brands as many tools as we can to help you build and maintain your structured product data. One of those tools is a bulk updater for product categories. We had one store use it with a product catalog of 13,000 products. They estimated it saved them a week’s worth of time if they had manually accepted the Magic category recommendations.
Shopify also uses Magic to recommend the values for category attributes. But those also must be accepted one at a time. Again Datify has a tool to bulk accept recommendations for category attributes. This tool saves even more time than the category recommendations as each product has only one category but many attributes.
Product Taxonomy in Existing Catalogs
Let’s say your store isn’t new, but has been around for a while. You probably have a legacy taxonomy based on unstructured product entities like product types and tags. You should migrate your legacy taxonomy to product categories and category metafields. Of course, every store will have set up it legacy taxonomy differently, so you’ll need to determine what legacy elements will be migrated where. That said, many stores will be migrating products types to product categories and tags to category metafields.
As these are common scenarios, Datify has tools for both.
The Product Migrator let’s you specify a metafield or category for each product type used in your store and then let you migrate the data in bulk.
The Tag Migrator does the same thing for migrating tags to metafields or product categories. Now, in legacy taxonomies, sometimes a bit of structure is added to tags by adding a prefix, like color_, to all tags that refer to color. The Tag Migrator has a feature that takes this tag taxonomy into account, which allows you to setup all of your migrations much more quickly.
Moving Forward
Now getting product categories and category metafields setup and populated is just Round 1 of this work. There’s going to more work in the future as Shopify continue to improve its Agentic sales channels.
Some of the work that we are doing for our clients that goes beyond Round 1 includes:
1. Working with Shopify to deepen the SPT. Shopify has a process where we can request:
a) new nodes be added to the product categories if the existing ones aren’t specific enough
b) new metafields be assigned to categories
c) new values be added a given metafield
2. Extending the category metafields beyond the SPT by adding our own category metafields if we cannot get Shopify to add them to the SPT.
3. Extending the metafield values beyond the SPT by adding our own values if we cannot get Shopify to add them to the SPT.
4. Adding this structured data to the product page’s JSON-LD data which helps with SEO.
And that last one brings up something that I want to mention. And that is good structured data for agentic also means good structured data for other purposes. So this structured data will help you with SEO and it will also help you with building better shopping experiences. For example you can use the structured data to have better filters and collection pages or spec tables on product pages.
Closing
My speculation is that AI will become an important channel for brands in the future. It won’t replace all other channels though it may replace some. But you’ll still be dependent on many of your existing channels. How important of a channel AI becomes will depend not on the technology, but on customer adoption.
The stores that will win in agentic e-commerce in 2026 are not the best designed. They’re not even the most branded. They are the most structured and consistent. I’m working with all of my clients to prepare them for this new sales channel. I recommend you get started on for your brand soon.