Preparing Retailers for Consumer Purchasing on ChatGPT and all other LLMs

As conversational AI and agentic commerce become increasingly prevalent, preparing product data for sale via ChatGPT and other LLMs requires a precise, data-driven approach.

In previous articles on Agentic commerce and leveraging ChatGPT, we discussed how retailers can improve their recommendability and citability. We are now focusing on the later stage of the buying funnel, where product purchasing journeys can be executed accurately within LLM systems.

The Trends:

Research in 2024–25 shows that roughly 35–40% of consumers use AI tools for product discovery, with usage exceeding 50% among Gen Z and Millennial shoppers. Among high‑frequency online buyers (those who make multiple purchases per week), about two‑thirds report using assistants like ChatGPT to guide purchase decisions, indicating that LLMs are already embedded in customers' habits.

However, traditional search and marketplaces still account for the majority of online purchasing volume, so LLM shopping is additive rather than fully substitutive at this time. But the writing is on the wall.

There is an agentic commerce opportunity, and for retailers to leverage this future, a high standard of structured data must be in place for LLMs to confidently recommend products for purchase. Two specific product-specific data disciplines stand out as the priorities in the short term...

1. Product variant structures
2. Product pricing, availability, and shipping

Product Variants: Defining "Option" and "Value" Product Dimensions

A product "option" is the various choices a consumer has for a specific product: things like size and colour. The product "values" is the particular breakdown of the options: red, green, blue, small, medium, large. The standard term for this is product variants and/or parent-child relationship.

Being precise in these relationships is crucial for presenting, validating, and selecting the correct SKU for consumers to purchase via ChatGPT.

The key is to structure the data so that LLMs can clearly identify each option and its value. This requires accurate technical mapping of SKUs to each product at the “child” level: for example, the SKU for a small red t-shirt.

This is a basic level of data preparation. Most retailers have a firm grip on this level of data structure.

However, these retailers have this level of data structure within their own ecosystems and, possibly, across marketplaces. The challenge is now to produce this layer of data integrity for LLMs. For many, this is a very different challenge.

Only once the above is set, can you then consider product bundles and kits.

Bundles and Kits:

LLMs require each bundle to have its own SKU and an additional layer of data to define it.

• List of SKUs that make up the bundle
  
• Other dependencies, such as inventory availability
• If there are variations to the bundle

Pricing, Inventory, and Shipping:

With the above validation markers in place for accurate product selection, the next step is to enable the LLMs to transact. Achieving this requires tightly integrating agentic checkout flows with product data, including pricing, inventory, and shipping logic.

Pricing:

ChatGPT needs a base price, promotional pricing, and bundle discounts if available.

It is essential to ensure that price calculation rules are deterministic.

Being "Deterministic":

The term “deterministic” is used intentionally; it means the activation of systemic rules to guarantee the same output for the same input. In other words, if a consumer repeatedly searches across different LLMs for your product, they will always see the same pricing, shipping methods, and costs.

LLMs require pricing data structures to achieve this level of reliability. Retailers who can accomplish this will be looked upon more favourably by the lies of ChatGPT when consumers are looking for your products.

Inventory Availability:

This is where your product master data links SKUs to real-time stock status, lead times, and order quantity limits.

Clearly indicate which SKUs are available for purchase in each region and surface the eligibility rules for LLMs to enforce at checkout.

Shipping:

Attaching shipping attributes (weight, dimensions, region eligibility, methods, costs) to each SKU, allows for deterministic shipping calculations. Again, a crucial part of the criteria needed for processing purchases on LLMs.

Structuring data for LLM Checkout and Purchasing Flows:

The ability to minimise purchasing ambiguity and reliably guide consumers through product selection, checkout, and payment in LLM-powered commerce channels is coming.

The more your data is perceived as "ambiguous", the less likely you will be at the forefront of this agency commerce future.

Ready to Grow? Let's Go? Together, we can work on your future in agentic commerce.