Agentic commerce: preparing products and prices for AI shopping agents

Abstract orange illustration on a dark background: a shopping cart. Article topic: “Agentic commerce: preparing products and prices for AI shopping agents”
In brief

Agentic commerce describes purchases in which an AI assistant searches, compares and, in some cases, completes the purchase on the user’s behalf. It is still emerging and capabilities vary, but preparation is concrete: complete product data, reliable prices and stock, clear policies and machine-readable pages.

The essentials in a definition

Agentic commerce is the use of AI assistants that search, compare and, with permission, buy on the user’s behalf. For the advertiser, it means product data, policies and pages must be complete and machine-readable, as asked for by Merchant Center, product structured data and Schema.org.

What changes for the advertiser

When an agent compares products, it decides based on structured data and text it can interpret, not just images and slogans. Products with incomplete attributes, inconsistent prices or vague policies risk being ignored or misdescribed.

Capabilities and formats depend on each platform and evolve fast. Consult current official documentation before planning any integration.

Complete product data

Work on the Merchant Center feed and product structured data: identifiers, brand, attributes, price, availability, shipping and returns. Keep what is in the feed identical to the page, and update price and stock changes without delay.

Describe relevant differences (compatibility, material, size, warranty) in plain, verifiable text, so an agent can distinguish similar products.

Illustrative diagram with 4 numbered steps joined by an orange line, the last one marked as done. It represents: Preparing the catalogue for shopping agents

Preparing the catalogue for shopping agents

  1. Complete attributes, identifiers and prices
  2. Make policies clear and consistent
  3. Mark up structured data faithful to the page
  4. Test how assistants describe the catalogue

Policies and trust

Publish shipping, returns, warranty and contact policies clearly and easy to find. An agent, and the customer using it, needs to know what happens if something goes wrong. Consistent information across the site reduces errors.

Consider the Schema.org product vocabulary to mark up offers, returns and genuine reviews, without exaggeration.

Large orange quotation marks on a dark background with abstract lines of text, illustrating the quote: “An agent only buys well what it can read without ambiguity.”
An agent only buys well what it can read without ambiguity.

Test with caution

Ask AI assistants about your products and categories and observe how they are described: what is right, what is missing, what is confused. Fix the source of the error in data or pages, instead of trying to manipulate the answer.

Keep current channels running. This trend is a complement, not a substitute, and measurement of results is still limited.

Decision table

ElementWhat the agent needsWhat to do
IdentificationBrand, GTIN and attributesComplete the feed
Price and stockCurrent, consistent valuesUpdate frequently
PoliciesClear shipping, returns and warrantyObjective, visible pages
ContentVerifiable differencesPlain, factual text

Common mistakes to avoid

  • Incomplete data. The product may be left out of comparisons.
  • Different prices in feed and site. It breeds distrust and disapprovals.
  • Betting everything on a trend. Capabilities are still changing.

How to measure the result

Track feed completeness, disapprovals, price consistency and the quality of descriptions in assistant answers, by periodic sampling.

Limits and exceptions

This is a recent trend, with capabilities, formats and rules that change and vary by market, and results are hard to measure. There is no guarantee that better data leads to more sales. Confirm current conditions in official documentation and invest proportionately.

Illustrative example

Illustrative example: a sports-equipment store audits its feed and finds brand and GTIN missing for part of the catalogue, prices misaligned with the site and a confusing returns page. It fixes the data and simplifies the page. When testing AI assistants, product descriptions become more accurate and complete.

Frequently asked questions

Is agentic commerce already a reality?

It is emerging, with capabilities that vary by platform and market; prepare the data, without betting everything.

Do I need to change the feed?

Usually, completing it and keeping it consistent is enough, following the current specification.

Can I influence what the assistant says?

You can improve data and content quality, but you do not control the answer.

What matters most?

Complete, consistent product, price, stock and policy data.

Should I create pages only for agents?

No; improve existing pages for people and machines at once.

Summary to quote

For agentic commerce, complete and align product, price, stock and policy data, mark up information faithfully and test how assistants describe the catalogue.

Sources and recommended reading

Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.

Keep reading

Quick contact

Want to talk about your campaigns?

Leave a short note. I'll reply within one working day.