Mar 19, 2026 | 4 minute read
written by Bryan House
Every new technology goes through a predictable cycle. Early excitement builds momentum, then the first signs of friction appear, and the conversation quickly shifts toward skepticism. We are in that phase again as OpenAI adjusts how transactions happen inside ChatGPT. Rather than completing purchases directly in the interface, transactions are moving back into merchant-controlled environments.
That shift triggered a familiar reaction. Some declared agentic commerce finished before it really started. Others argued merchants would never allow external platforms to sit inside the transaction layer.
If you zoom out, none of this is surprising. Early market behavior often looks messy because it is messy. The first version of a new model rarely survives unchanged.
The current narrative suggests consumers do not trust AI enough to complete transactions inside an interface like ChatGPT. That explanation feels intuitive, but it misses the real issue.
Consumers already complete transactions in relatively new environments. Just look at social commerce platforms like TikTok Shop, which generated $33B in GMV in 2024, including about $9B in the U.S. alone. Three years ago the platform was doing roughly $1B. Adoption followed once the experience became reliable and predictable.
The constraint here is not trust. It is execution.
Completing a transaction depends on accurate pricing, inventory availability, tax calculation, fulfillment logic, and clear product definition. When any of those elements break, the experience fails. AI interfaces surface those gaps more quickly because they depend on structured inputs rather than predefined flows.
This is where the problem shifts from interface design to underlying data.
If the underlying product data is incomplete or inconsistent, no checkout model will perform reliably, whether it lives inside an AI interface or a merchant application.
While most of the recent agentic commerce discussion focuses on checkout mechanics, the real constraint sits somewhere else entirely: product data. Even if trust were solved overnight, most catalogs would still not support agent-driven transactions.
If AI agents are going to shop on behalf of customers, they need structured information they can interpret without guessing. Unfortunately, most commerce catalogs were never designed for machines. They were designed for humans browsing websites.
Descriptions are inconsistent. Attributes vary from product to product. Important details live in PDFs, spreadsheets, or marketing copy written for a catalog page rather than a machine. An AI agent cannot reason about a product it cannot understand.
Consider a product description that says something like, “Industrial pump. High performance. Suitable for multiple environments.”
That sentence might work for a marketing page. It does not help a machine compare specifications across suppliers or determine whether the product meets a specific technical requirement. Machines need structured attributes, consistent taxonomy, and metadata that describes exactly what the product is, how it works, and how it relates to other products.
Without that structure, agentic commerce simply cannot function.
Product catalogs have historically been treated as supporting systems. They powered the storefront, but they rarely shaped strategic decisions. That role is changing.
In an agent-driven environment, the catalog becomes the interface between merchants and machines. It is the dataset agents rely on to discover products, evaluate options, and recommend purchases. When that data is structured and consistent, agents can reason effectively. When it is fragmented, products become harder to surface and harder to compare.
This shift is already visible in early protocol development. Efforts like Google’s Universal Commerce Protocol and emerging approaches from companies such as Stripe are attempting to define how systems interact in an agentic ecosystem. Protocols tend to appear early because they give developers a common foundation to build on. Over time, they help establish shared standards across platforms.
Those standards ultimately depend on structured product data. Without that layer, interoperability remains limited.
For years, teams treated the product catalog as content. They added descriptions, uploaded images, and configured a limited set of attributes so products could render on a page. Once published, the catalog changed infrequently.
Agentic commerce changes that expectation.
As the catalog becomes more central to how commerce operates, product data starts to resemble software. Teams define schemas that enforce consistency. Attributes remain standardized across product lines. Relationships between products are modeled explicitly. Updates are versioned and governed so downstream systems and agents can rely on them.
The catalog begins to behave like code.
The industry will continue to experiment with transaction models and protocols. That process will take time. The more meaningful shift is happening beneath the surface. AI agents will favor catalogs that are structured, consistent, and easy to interpret. Those catalogs will determine how products are discovered, compared, and ultimately purchased.
Schedule a demo to see how Elastic Path delivers unified commerce for leading global brands.