The categoryUpdated 2026-08-17

AI commerce: which parts change what you have to publish

AI commerce covers everything from recommendation engines and demand forecasting to generated product copy and AI shopping agents. Most of it is internal tooling: it changes how you operate. One part changes what the outside world can read from your store — and that is the part with an external dependency you cannot fix later.

Definition

AI commerce is the application of machine learning and generative models across the ecommerce value chain, spanning internal operations such as merchandising, pricing and support, and external demand mediated by AI shopping agents.

The map

Grouping AI commerce by who consumes the output makes the priority obvious. Internal applications improve margins on demand you already have. Agent-mediated demand determines whether you are in the consideration set at all.

AreaWho consumes itFailure mode if ignored
On-site recommendationsYour existing visitorsLower AOV; recoverable at any time
Demand forecasting and pricingYour operations teamMargin erosion; recoverable
Support automationYour existing customersHigher cost per ticket; recoverable
Generated content and imageryYour existing visitorsSlower catalogue velocity; recoverable
Agent-mediated demandAI shopping agents acting for buyers who never reach your siteSilent exclusion from the consideration set; invisible in analytics
AI in ecommerce, by consumer of the output

Why the last row is different

Every other line in that table fails loudly. Conversion drops, tickets pile up, margin compresses — you find out, and you can act.

Agent-mediated demand fails silently. There is no referrer for a comparison you lost inside a model's context window, no impression count for a candidate set you were excluded from, and no error when an agent fetches your product page and finds no product data in it. In our 120-store sample, 64% would fail that fetch today.

The work overlaps with what you already do

The readiness fixes are not a separate programme. Complete Product structured data improves rich results and shopping feeds. A resolvable product sitemap improves indexation. Machine-readable return and delivery terms reduce pre-purchase support contacts.

This is the practical argument for doing the external-facing work first: it is the only category with an irreversible failure mode, and it pays for itself through channels you already measure.

Structured data
Serves agents, Google rich results, Merchant Center and comparison sites from one source.
Crawl access
Serves retrieval agents and classic search indexation with the same robots rules.
Policy data
Serves agent risk resolution and human pre-purchase questions with the same fields.

Frequently asked questions

What is AI commerce?
AI commerce is the application of machine learning and generative models across the ecommerce value chain — merchandising, pricing, support, content generation — plus the external demand now mediated by AI shopping agents.
What is the difference between AI commerce and agentic commerce?
AI commerce is the umbrella, most of which is internal tooling. Agentic commerce is specifically the demand side: transactions where an AI agent acts for the buyer. Agentic commerce is the subset with an external dependency on what your store publishes.
Which AI commerce investment should a small merchant make first?
The external one. On-site recommendations and support automation improve demand you already have and can be added at any point. Being unreadable to AI shopping agents removes you from demand you never see, and the fix is cheap: structured data, crawl access, machine-readable policies.
Does AI-generated product copy help with AI shopping agents?
Marginally. Agents read fields, not prose. A generated description in an unstructured page is still an unparseable product; the same description inside a complete Product JSON-LD block is a usable record.

Start with the part that fails silently

Scan your store and see whether agent-mediated demand can reach you at all.

Your store vs. competitor

Unlocks AI preference

Free scan · ~30 seconds · No account

Related

From the blog