The way people search for products is changing. Consumers are already using artificial intelligence tools to compare options, narrow down choices by price, features, and availability, and find the product best suited to a specific need.
For retailers, this creates a new challenge: having the right product is no longer enough. You must ensure that the information representing that product is complete, up-to-date, and connected across different channels.
This guide shows what to review in your operations to increase the likelihood of your products being understood and found in AI-mediated search and shopping experiences.
1. Start with catalog quality
Before thinking about artificial intelligence, it is worth looking at a more basic question: do your own systems clearly understand what you are selling?
A structured catalog must allow for the identification of not just the product's trade name, but also its specific features.
Consider a pair of sneakers.
"Women's Run Pro Sneakers" might be enough for someone who already knows that model. But a tool trying to answer a question like "which waterproof women's sneakers under $700 are available near me?" will need to correlate several different pieces of information (brand, model, category, gender, color, etc.).
The better these attributes are structured, the easier it becomes to distribute the same information to search engines, marketplaces, ad campaigns, AI platforms, and other channels.
What to review
Take a sample of your best-selling products and check:
- Do titles follow a standard format?
- Are variants correctly separated?
- Are important attributes filled out?
- Do products have consistent identifiers?
- Do descriptions explain actual features, or are they just promotional copy?
- Does the same product have different information in different systems?
Start with the categories that generate the most revenue or have the highest search volume and identify patterns of inconsistency.
2. Define a single source of truth for every piece of information
In larger operations, it is common to find different versions of the same product scattered throughout the company.
The ERP shows one price. The e-commerce site shows another. A physical store has already sold the last unit, but the inventory still appears available online. The marketplace received the update a few hours ago.
The problem is that new discovery experiences depend precisely on this information.
ChatGPT, for example, already uses public data and information provided directly by merchants to discover and compare products. When choosing between retailers offering the same item, availability, price, and quality are among the factors considered.
This makes it important to answer an operational question:
Which system represents the most reliable version of each piece of data?
You might have, for example:
PIM: product information and attributes
ERP: commercial pricing
OMS: sales availability
WMS: physical stock location
POS: store transactions
e-commerce: consumer presentation
The design varies from company to company. What matters is that the data has a known source and a clear update logic.
Create a simple map
Choose five fundamental pieces of information:
Product → Price → Inventory → Store → Availability
For each one, answer:
- Where does this information originate?
- Which system is considered the official source?
- Which channels receive the data?
- How frequently is it updated?
- What happens when systems diverge?
This analysis often reveals issues that don't appear when each department only looks at its own platform.
3. Keep price and inventory truly up to date
Institutional information can remain valid for months. Price and inventory can change in minutes. This difference matters significantly for systems intended to help the consumer make a purchasing decision.
OpenAI already allows retailers to provide product feeds directly to improve the updating and representation of their catalogs in ChatGPT. The Agentic Commerce Protocol also provides for increasingly complete product information and, in the future, local availability and delivery estimates.
Google is moving in the same direction. Updates to the Universal Commerce Protocol allow agents to query details in real time, including price and inventory, directly from retailer catalogs.
See the Universal Commerce Protocol updates
This doesn't mean every retailer needs to implement these protocols immediately. It means that the ability to provide up-to-date commercial information is gaining importance.
4. Structure availability by store
For retailers with physical locations, there is another level of complexity.
It is not enough to answer: “Do we have this product in stock?”
You need to be able to answer: “Which store has this product in stock?”
This information is already used in traditional discovery. Google’s free local listings allow you to show products available in physical stores to nearby consumers, including across experiences like Search, Maps, and Business Profile.
Learn about Google’s free local listings
To enable local inventory features, Google requires linking stores to Merchant Center and submitting the necessary inventory information. This provides a good benchmark for assessing operational maturity.
5. Make your products machine-readable
A visually perfect page does not necessarily mean a page that is easy for automated systems to interpret. Beyond the content displayed to the consumer, there are structured ways to communicate what that page represents.
In the Google ecosystem, structured product data can provide information on elements like price and availability. Google also recommends combining structured data on your site with information submitted to Merchant Center whenever possible.
But there is an important caveat: do not treat structured data as a list of keywords to fill in. It must accurately represent what is actually on the page and in your operations.
6. Prepare your feeds for new discovery channels
For years, many retailers thought of product feeds almost exclusively as infrastructure for Google Shopping or marketplaces.
That perspective is becoming outdated.
The feed is becoming a way to distribute your catalog to different discovery environments. In March 2026, OpenAI expanded the Agentic Commerce Protocol for product discovery and began allowing merchants to share feeds to represent their catalogs in ChatGPT. Major retailers like Target, Sephora, Nordstrom, Lowe's, Best Buy, The Home Depot, and Wayfair were already integrated into the protocol when the initiative was announced.
See how product discovery works in ChatGPT
Evaluate your feed
Pay special attention to:
- unique identifier;
- brand;
- title;
- description;
- category;
- category-specific attributes;
- variants;
- images;
- price;
- promotions;
- availability;
- URL;
- store-specific information, where applicable.
And find out how many transformations the data undergoes before reaching the final channel.
The more spreadsheets, manual adjustments, and independent integrations there are between the source and the destination, the higher the risk of inconsistency tends to be.
7. Test how your products are being found
GEO for retail should not be treated merely as content production.
It is worth creating a discovery audit routine. Choose between 10 and 20 real search intents related to the company's products.
For example: “Men's running shoes under $500”, “55-inch OLED TV with fast delivery”, “Automatic coffee maker available near me”, and/or “Laptop for video editing under $8,000”.
Perform these searches periodically on traditional search engines and AI tools that offer product discovery, and record:
- Which brands appear;
- Which products appear;
- Which attributes are highlighted;
- Whether your company appears as a retailer;
- If price and stock are correct;
- Which competitors appear;
- Which sources are used;
- Where information seems to be incomplete.
The initial goal shouldn't be to create an "AI ranking." There is no single standard for positioning across these platforms yet. The goal is to identify gaps in catalog representation.
Turn tests into actions
If products aren't appearing, investigate:
Is information missing?
Perhaps relevant attributes are absent.
Is the information inconsistent?
Check feeds and integrations.
Does the product appear, but another store is prioritized?
Compare price, availability, and the quality of commercial information.
Does the AI find the product but not the nearby store?
Review the local inventory structure.
Are competitors frequently cited as a reference?
Analyze what kind of information they provide that your brand does not.
This turns GEO from an abstract discussion into a concrete diagnostic routine.
8. Before investing in GEO for products, review these 10 questions
- Do we have a consistent identifier for each product and variant?
- Are the essential category attributes filled out?
- Are prices and promotions updated automatically?
- Do we know the availability per store?
- Does the digital inventory match the actual stock?
- Does our website present structured data correctly?
- Are Merchant Center and other feeds up to date?
- Do we know which system is the official source for each piece of information?
- Can we distribute the catalog to new channels without creating new manual integrations?
- Do we regularly test how our products appear in search results and AI tools?
If you answered "no" to several of these, there is likely infrastructure work to be done before considering advanced optimization.
What changes from here on out
The main change isn't that consumers will stop searching on Google or visiting online stores. It's that discovery is now happening across more interfaces.
A person might start on ChatGPT, compare on Google, visit a store, complete the purchase in an app, or pick up an order a few hours later. In this scenario, the product must maintain its identity throughout the entire journey.
Therefore, preparing products for AI is less about "optimizing for an algorithm" and more about building a commercial database that can keep up with new discovery channels.
