Skip to content

AI Agents

How brands get chosen in agentic commerce

AI agents now compare and buy products for people. See how they choose, and what gets your brand picked.

Free agent readiness scan · no credit card

Agent compares · 10 seats

yourbrand$12 / seat
competitor$15 / seat
other.io$18 / seat

Order placed

yourbrand CRM
10 seats$120 / moPaid by agent✓

Agentic commerce meaning

Agentic commerce means AI agents that find, compare and buy products for a person or business, inside limits that person sets. IBM defines it as agents that research, negotiate and complete purchases, often without direct human intervention. It runs on open checkout protocols and tokenized agent payments, and it rewards brands whose product facts agents can read and trust.

01 · Side by side

Agentic commerce vs traditional e-commerce.

Traditional e-commerce is built for a person clicking through pages; agentic e-commerce is built for software reading data. Signifyd sums up the shift: the agent prepares the cart and does the manual steps, and people step in at approval and payment.

Who browses
Traditional e-commerceThe shopper, page by page
Agentic commerceAn AI agent, across many sites at once
First touchpoint
Traditional e-commerceSearch results, ads, your homepage
Agentic commerceThe agent's answer or shortlist
What persuades
Traditional e-commerceDesign, copy, reviews, offers
Agentic commerceStructured facts: price, stock, specs, policies
Checkout
Traditional e-commerceShopper enters details on your site
Agentic commerceAgent pays with a tokenized, approved credential
Fraud checks
Traditional e-commerceBased on human session behavior
Agentic commerceMust tell trusted agents from malicious bots
What to measure
Traditional e-commerceSessions and conversion rate
Agentic commerceHow often agents mention, cite and pick you

02 · Definition

What is agentic commerce?

Agentic commerce is a way of buying and selling in which AI agents do the shopping work. A person or a business states a goal, and an agent researches, compares, negotiates and, inside limits the person sets, completes the purchase. IBM defines it as AI agents acting on behalf of consumers or businesses to research, negotiate and complete purchases, "often without direct human intervention."

Stripe describes the same thing from the shopper's side: AI agents finding, comparing and potentially making purchases for customers. The word "potentially" matters. Most agents today still ask a person to approve the final step, and the level of autonomy is set by the shopper, the platform and the payment method.

The difference from earlier shopping assistants is action. A recommendation engine suggests a product and hands the task back. An agent carries the task forward: it fills the cart, checks delivery, and pays with a credential issued for that purpose.

The terms overlap. Agentic e-commerce, agentic shopping and agentic AI commerce all describe the same shift: software agents doing the finding, comparing and buying that people did by hand. The definition used by most sources is the one above.

1yourbrand
2
3

03 · The journey

How does agentic commerce work?

Agentic commerce works as a journey from discovery to transaction, run by an agent in six stages. IBM and Stripe describe the same sequence: a stated need, a search, an evaluation, a decision, a secure payment and support after the sale.

  1. 1

    Intent

    The shopper gives a goal, a budget and rules, such as "white running shoes under $150, delivered by Friday."

  2. 2

    Discovery

    The agent searches the web, retailer catalogs and product feeds. Stripe notes agents often scan options through the Model Context Protocol, an open standard for connecting AI apps to outside systems.

  3. 3

    Evaluation

    It compares price, stock, delivery dates, reviews and return policies across sellers. This stage decides which brands are considered at all.

  4. 4

    Decision

    It picks an option, or shows a short list when the rules require a person to approve.

  5. 5

    Checkout and payment

    It pays through a protocol and a tokenized credential. In Google's Agent Payments Protocol, a Cart Mandate records the exact items and price before money moves.

  6. 6

    After the sale

    It tracks the order and handles returns or support questions.

04 · Infrastructure

Which protocols power agentic commerce?

Open protocols let agents and merchants talk in a shared format, so an agent can check out on a store it has never seen. Payment networks add tokens and signatures so a merchant can trust the agent and the person behind it. These are the main standards in use.

Agentic Commerce Protocol (ACP)
Backed byStripe and OpenAI
What it doesOpen standard, Apache 2.0, for commerce flows between buyers, AI agents and businesses. ChatGPT was the first platform to implement it.
Agent Payments Protocol (AP2)
Backed byGoogle, with 60+ organizations
What it doesSigned mandates that prove what a shopper approved. Announced September 16, 2025.
Universal Commerce Protocol (UCP)
Backed byGoogle with Shopify, Etsy, Wayfair, Target, Walmart
What it doesCovers discovery, buying and post-purchase support. Checkout is live in AI Mode and the Gemini app for eligible US retailers.
Visa Intelligent Commerce and Trusted Agent Protocol
Backed byVisa, with Cloudflare for TAP
What it doesAgent payments, plus cryptographic signatures so merchants can tell trusted agents from bad bots.
Mastercard Agent Pay
Backed byMastercard
What it doesAgentic Tokens built on Mastercard's existing card tokenization. Announced April 29, 2025.

05 · Payments

How agentic commerce payments work.

The common thread is that an agent never needs the raw card number. Most merchants will get this through their payment provider rather than build it.

  1. 01

    Shared tokens

    Stripe issues a Shared Payment Token for ACP checkouts, so card details stay hidden from the agent.

  2. 02

    Network agent tokens

    Mastercard Agent Pay and Visa Intelligent Commerce issue credentials built for agents on top of existing card tokenization.

  3. 03

    Wallet programs

    PayPal's Agent Ready lets existing PayPal merchants accept payments on AI surfaces.

06 · In the wild

Agentic commerce and agentic shopping examples today.

Agentic commerce is live in several large products, though most still keep a person in the approval step.

  • 01

    Google AI Mode and Gemini

    UCP checkout works for eligible US retailers, and Business Agent is live for brands including Lowe's, Michaels, Poshmark and Reebok (Google, January 2026).

  • 02

    Shopify Agentic Storefronts

    Merchants sell inside ChatGPT, Copilot, AI Mode and Gemini while staying merchant of record (Shopify, March 2026).

  • 03

    Meta Muse

    Launched in the US on September 8, 2026, Muse can check out with Link by Stripe and checks with the person before buying (Meta).

  • 04

    Visa pilots

    Partners completed hundreds of agent-initiated transactions in controlled real-world settings (Visa, December 2025).

  • 05

    A launch that pulled back

    ChatGPT's in-chat Instant Checkout started in September 2025. By March 2026 Forrester reported OpenAI had pulled it back, sending checkout to merchant apps and sites instead.

07 · Why it matters

What are the benefits of agentic commerce?

The main benefit is time: agents take over the searching, comparing and form filling that shoppers did by hand. For businesses, the benefit is reach into a channel where more buying decisions now start.

  1. 01

    For shoppers

    Less tab switching and manual comparison. Routine and repeat purchases handled inside set limits, with checks on price, stock and delivery across many sellers.

    1yourbrand
    2
    3
  2. 02

    For businesses

    A channel to shoppers who may never visit your site, fewer abandoned checkouts, and automated B2B sourcing and reorders, which IBM lists as a leading use case.

    AI mentions

    1yourbrand
    2
    3
  3. 03

    For the market

    Forecasts vary widely. McKinsey, cited by IBM, puts agentic commerce at $3 trillion to $5 trillion globally by 2030; Morgan Stanley estimates up to $385 billion of US e-commerce.

    AI mentions

    1yourbrand
    2
    3

08 · Risks and trust

What are the risks, and how is trust built?

The biggest open problem is trust: a merchant needs to know the agent is legitimate, that a real person stands behind it, and that this specific purchase was authorized. A Veriff-sponsored piece in Retail Dive puts it plainly: a standing order to reorder groceries is not permission to ship electronics to a new address.

  • Agent verification

    Telling buying agents from scrapers and fraud bots. Visa's Trusted Agent Protocol and the Know Your Agent work from Visa, Mastercard and Ant International target this.

  • Liability

    JPMorgan told Payments Dive the rules "are not fully formed yet." American Express says it will protect eligible cardholders from charges caused by AI agent error when purchase intent is authenticated (Payments Dive).

  • Consent

    In an RTB House survey, 35% of shoppers wanted a person to review transactions before an agent buys (Retail Dive).

  • Privacy

    IBM research found 83% of consumers worry about privacy, data misuse and unwanted marketing.

  • Platform access

    Stores decide which agents may shop. Amazon blocked Meta's Muse, saying access by an unauthorized AI agent violates its terms (GeekWire).

09 · Checklist

How brands and marketplaces can prepare for agentic commerce.

Prepare by making your products easy for an agent to find, read, trust and buy. IBM, Salesforce and Stripe give the same advice in different words: machine-readable data first, clear rules second, agent access third.

  1. 1

    Complete product data

    Price, stock, variants, materials and delivery times, matching across your site, feeds and marketplaces. OpenAI's product feed spec requires fields like item_id, title, price and availability, and asks merchants to keep price and availability current.

  2. 2

    Structured data

    Product, Offer, shipping and return policy markup, using schema.org and GS1 standards as Salesforce recommends.

  3. 3

    Clear policies

    Returns, warranty and delivery terms on crawlable pages, written as plain facts an agent can quote.

  4. 4

    Comparison content

    Honest pages that say who a product suits and how it differs from alternatives.

  5. 5

    Third-party proof

    Reviews and mentions on sites that AI assistants read and cite.

  6. 6

    Agent access

    Bot rules that block abuse without blocking verified buying agents.

  7. 7

    Protocol readiness

    Ask your commerce platform and payment provider which of ACP, AP2, UCP and network agent tokens they support.

  8. 8

    Measurement

    Track how often ChatGPT, Gemini, Perplexity and Google AI Overviews mention and cite your brand, and for which prompts.

  9. 9

    Verify the catalog

    Verify sellers and listings so agents can trust the catalog, then expose inventory through the feeds and protocols agents already use, such as Google Merchant Center's conversational attributes and Shopify Catalog.

10 · Straight answer

Why AI visibility comes before agentic checkout.

An agent can only buy what it first recommends, and it recommends what it can find and cite. That makes AI visibility the leading signal for agentic commerce.

  1. 01

    Checkout decides how an agent pays

    Visibility decides whether your product reaches the shortlist at all.

  2. 02

    Agents read the same web assistants cite

    IBM names generative engine optimization as part of preparing: structuring product content so models and agents can interpret it.

  3. 03

    It can be measured today

    Track how often ChatGPT, Gemini, Perplexity and Google AI Overviews mention and cite your brand, then fix the pages and facts they skip.

11 · Latest

Latest developments and the future of agentic commerce.

Reviewed weekly while agentic commerce is in the news.

  1. 1

    Anthropic publishes a commerce agents blueprint

    Claude for commerce agents is an open blueprint for shopping and merchant agents. Payment is left to the builder's checkout or payments provider.

  2. 2

    Meta launches Muse

    Muse launched in the US, with checkout through Link by Stripe.

  3. 3

    Card networks start a Know Your Agent framework

    Ant International, Mastercard and Visa began work on interoperability covering operator traceability, shared certification and monitoring.

  4. 4

    Amazon blocks Muse

    GeekWire reported Amazon blocking Muse as an unauthorized agent.

12 · FAQs

Common questions about agentic commerce.

What is the difference between agentic AI and agentic commerce?

Agentic AI is the technology: AI systems that pursue a goal with limited supervision. Agentic commerce is one use of it, where the autonomy is focused on shopping and buying, as PayPal puts it.

Can you give an example of agentic commerce?

A shopper asks Google's AI Mode or Meta's Muse for a product. The agent compares options, prepares the cart and, after the shopper approves, checks out with a tokenized payment such as Google's UCP checkout or Link by Stripe.

Which companies are using agentic commerce?

Google, Shopify, Meta, Stripe, OpenAI, PayPal, Visa and Mastercard all run agentic commerce products or protocols. Retail brands such as Lowe's, Michaels, Poshmark and Reebok use Google's Business Agent.

Is agentic commerce the next big thing?

It is early. Protocols and payment programs are live, while adoption, liability and platform access rules are still forming. Stripe's founders say the industry is at the first levels of autonomy.

Is agentic AI just hype?

Some forecasts run ahead of reality, and OpenAI scaled back in-chat checkout in 2026. But the core pieces are real: agents that act, open checkout protocols and card network tokens for agent payments.

How do brands show up in agentic commerce?

By being findable and citable. Agents shortlist products they can read and verify, so complete product data, structured markup, clear policies and mentions on trusted sites raise the odds of being picked.

Related topics

Sources

  1. 01What is agentic commerce? · IBM, January 23, 2026
  2. 02What is agentic commerce? A guide to getting started · Stripe, accessed September 27, 2026
  3. 03Agentic Commerce Protocol · Stripe and OpenAI, accessed September 27, 2026
  4. 04Announcing Agent Payments Protocol (AP2) · Google Cloud, September 16, 2025
  5. 05Agentic commerce tools and the Universal Commerce Protocol · Google, January 11, 2026
  6. 06Visa Intelligent Commerce · Visa, April 30, 2025
  7. 07Trusted Agent Protocol · Visa, October 14, 2025
  8. 08Mastercard unveils Agent Pay · Mastercard, April 29, 2025
  9. 09PayPal launches agentic commerce services · PayPal, October 28, 2025
  10. 10Agentic commerce momentum · Shopify, March 24, 2026
  11. 11What it means that the leader in agentic commerce just pulled back · Forrester, March 7, 2026
  12. 12Agentic commerce vs traditional ecommerce: 7 key differences · Signifyd, December 19, 2025
  13. 13Agentic AI raises liability issues · Payments Dive, October 28, 2025
  14. 14Amex pushes for agentic standards · Payments Dive, April 14, 2026
  15. 15Stripe's slower view of agentic commerce · Payments Dive, February 25, 2026
  16. 16Retail shoppers warm up to agentic AI purchases · Retail Dive, August 13, 2026
  17. 17Preparing your platform for agentic commerce (sponsored by Veriff) · Retail Dive, September 8, 2026
  18. 18Morgan Stanley: agentic shoppers and online sales · Digital Commerce 360, December 9, 2025
  19. 19Product feed spec · OpenAI, accessed September 27, 2026
  20. 20Claude for commerce agents · Anthropic, September 2, 2026
  21. 21Introducing Muse · Meta, September 8, 2026
  22. 22Know Your Agent interoperability collaboration · Ant International, September 10, 2026
  23. 23Amazon blocks Meta's Muse AI assistant · GeekWire, September 21, 2026
  24. 24Visa agentic commerce pilots · Visa, December 18, 2025
  25. 25Model Context Protocol introduction · Model Context Protocol, accessed September 27, 2026