Agent commerce

A new shopping entry point is taking shape in conversation

A buyer states a need. An agent reads available product data, compares the options, and forms a shortlist. Discovery, selection, and the path to purchase now converge in a single conversational interface.

Agent commerce is a commercial path in which agents take part in product discovery, comparison, selection, and transaction execution.

Shopping now has a conversational entry point

Search and marketplaces remain central. Agents now organize another route into the consideration set.

Search

A buyer enters keywords and continues from a results page.

Marketplaces

A buyer enters a catalog or recommendation feed and browses a broad assortment.

Agent commerce

A buyer states a need. An agent organizes the available information, compares products, and returns a shortlist.

Brand competition now extends from earning the click to making the agent's shortlist.

Platform progress

Agent commerce has entered the infrastructure-building phase

Since September 2025, platforms have introduced product protocols, payment authorization, shopping surfaces, and merchant integrations. The paths are still diverging. The investment is already under way.

  1. September 2025

    Google introduced AP2, using signed mandates to define authorization for agent payments.

    Google Cloud

  2. September 29, 2025

    OpenAI and Stripe introduced the Agentic Commerce Protocol (ACP) and launched Instant Checkout.

    OpenAI Developers

  3. January 11, 2026

    Google and Shopify introduced Universal Commerce Protocol (UCP) for commerce across agent surfaces.

    Google Developers · Shopify

  4. January 2026

    Microsoft launched Copilot Checkout, extending commerce into its AI assistant.

    GeekWire

  5. March 2026

    OpenAI retired the initial Instant Checkout model, kept merchant checkout, and shifted its focus to product discovery.

    Retail Dive, quoting OpenAI

  6. May 11, 2026

    Qwen and Taobao connected product recommendations, ordering, fulfillment, and after-sales service.

    Xinhua News Agency

  7. May 2026

    Doubao introduced Help Me Choose, connecting product discovery to Douyin ecommerce.

    TMTPost

Protocols are separating by function

ACP, UCP, and AP2 address product exchange, commerce coordination, and payment authorization. Their boundaries are still evolving.

Discovery and checkout are separating

An early checkout model can change while platform investment in product discovery continues.

Product data is the shared foundation

Open protocols and integrated platforms follow different routes. Both depend on accurate, retrievable, and comparable product information.

The result is a shortlist

Search results can continue across pages. An agent response narrows the field before the buyer sees it. A product that misses the shortlist has no route into the next comparison or purchase step.

  • AI
    Long-tail discovery

    Find me a portable pour-over coffee kit for camping.

    AI
    • Portable Pour-Over Kit V2$49
    • Mini Hand Grinder Pro$39
    • Folding Drip Kettle$29
    Your brand isn’t here.
  • AI
    Known-item price check

    Where can I get the best price on an iPhone 15 Pro 256GB, new with warranty?

    AI
    • Apple Storefree next-day$1,099
    • Amazon2-day shipping$1,049
    • Best Buystore pickup$1,079
    Your brand isn’t here.
  • AI
    Scenario picks

    I need a thin-and-light laptop for office work and video calls, under $800, with long battery life.

    AI
    • Lenovo IdeaPad Slim 5$649long battery
    • HP Pavilion Aero 13$749ultralight
    • ASUS Zenbook 14$699balanced
    Your brand isn’t here.

At the new entry point, the first competition is for a place on the shortlist.

From index to action

An answer is the last step, not the first. Product data must first become retrievable, then support candidate selection, comparison, and a working purchase path.

  1. 1

    Index

    Web pages, feeds, and catalogs enter the sources available to the agent.

  2. 2

    Candidate set

    The agent forms an initial product set from the buyer's requirements.

  3. 3

    Live data

    Price, availability, delivery coverage, and purchase URLs remain retrievable.

  4. 4

    Comparison and presentation

    Specifications, use cases, reviews, and policies support the selection.

  5. 5

    Action

    The buyer continues on the merchant's site, or a supported agent carries the transaction forward.

Each layer depends on the one before it. Missing information changes the shortlist and every purchase step that follows.

Transactability

Discovery must connect to a working purchase path

Transactability is a product's ability to move from discovery through comparison and selection to a working purchase path. Missing price, unverifiable availability, incomplete specifications, or an unreachable purchase URL can break that path.

Discovery
Comparison
Selection
Purchase path

Recommendation opens the path. Transactability keeps it moving.

Why now

The channel is still forming. The operating signals are already here.

Current behavior
20%
AI and agents influenced 20% of purchases worldwide during Cyber Week 2025.
Salesforce · December 2025
Merchant relevance
75%
Shopify's Q2 2026 results showed that 75% of AI-attributed purchases occurred in categories outside the platform's top 100.
Shopify · Q2 2026
Long-range forecast
$3–5T
McKinsey estimates a $3–5 trillion global opportunity for agent commerce by 2030.
McKinsey · 2025
Data sources · Salesforce 2025 · Shopify 2026 · McKinsey 2025

Agent commerce is still taking shape, while search and marketplaces remain central. Brands now have an additional operating task: make product data retrievable, intelligible, and comparable to agents, then preserve a continuous record from the first run.

Product data foundation

Make the catalog retrievable in a form agents can interpret and compare.

Measurement baseline

Preserve the first run under a consistent method so later results have a valid point of comparison.

Correction record

Carry known issues, corrective action, and retest status into later cycles instead of restarting from zero.

The sooner you enter the new channel, the sooner your advantage starts to build.

Making the shortlist is only the beginning

Build a machine-readable catalog

Keep product identity, specifications, price, availability, use context, and purchase URLs accurate, current, and structured for agent retrieval.

Test continuously and fix at the source

Retest shortlists and purchase paths under the same method. Trace errors and gaps to the catalog, then check them again in a later cycle. The earlier you connect, the longer the record behind every proposed fix.

Getting in decides whether you appear. Continuous optimization decides how far you go.

SoldByAI brings both capabilities into one operating system, from catalog connection and weekly sweeps across the leading AI engines to purchase-path testing and engine-level attribution.

See whether your product data is ready for the new channel

Enter your store URL to see which product identifiers, structured data, and key signals are missing.

Free · No credit card required