Agentic commerce is shopping carried out partly or fully by AI agents acting for a person: researching options, comparing products, and in some cases completing the purchase. Agents rely on structured, accurate product data — feeds, schema, prices, stock, shipping and return policies — plus reviews and brand trust. To prepare: keep product feeds and schema complete and consistent, publish machine-readable policies, earn genuine reviews, and follow the checkout standards emerging from OpenAI, Google and payment providers.
On this page
- What is agentic commerce?
- The protocols behind agentic checkout
- How AI agents choose products
- Making product data readable by AI
- Where agentic shopping happens
- Agentic commerce for retailers and B2B
- Agent-ready ecommerce checklist
- Agent-friendly checkout
- Marketplaces and shopping assistants
- Agentic commerce risks
- The future of ecommerce
- Frequently asked questions
Search interest in agentic AI has grown quickly — Ahrefs has reported it at around 111,000 monthly searches, up strongly year on year, with “agentic commerce” rising too. The shift is practical: assistants such as ChatGPT, Google’s AI Mode and Perplexity can now compare products in depth, and some can complete purchases. This guide explains what is agentic commerce and how to prepare. For making your whole website agent-ready, see our agentic search guide.
What is agentic commerce?
Traditional ecommerce: a person searches, browses, compares and checks out. Agentic commerce: the person states a goal (“find me waterproof hiking boots under $150 that arrive by Friday”) and AI shopping agents do the research, shortlist options and, with permission, buy. AI shopping assistants today mostly help with research and comparison; full agentic checkout is growing as standards mature.
The protocols behind agentic checkout
- OpenAI introduced Instant Checkout in ChatGPT (often searched as instant checkout ChatGPT) with the Agentic Commerce Protocol (ACP), developed with Stripe, so merchants can let ChatGPT complete purchases.
- Google introduced the Agent Payments Protocol (AP2), an open standard for AI agent payments that records user authorisation for agent-initiated purchases.
- Platforms and payment providers continue to add support, and further standards are emerging — expect this area to keep changing.
Most merchants will access these through their ecommerce platform and payment provider rather than building them directly.
How AI agents choose products
How AI agents choose products comes down to whether they can understand and trust what you sell:
- Complete product attributes that match the request (size, material, compatibility, dimensions).
- Inventory and pricing accuracy — agents avoid products that may be out of stock or priced inconsistently.
- Clear shipping times, costs and a machine-readable return policy.
- Reviews and AI shopping go together: ratings and review content inform recommendations.
- Brand trust AI agents can verify — consistent information across your site, marketplaces and listings.
Making product data readable by AI
Good product data for AI is the same data that powers shopping results today, held to a higher standard of completeness and accuracy.
AI readable product data is the foundation. Practical steps:
- Product feed optimization: complete titles, descriptions, categories and attributes in your Merchant Center feed and any marketplace feeds — a product feed for AI agents is the same feed, kept rigorously accurate.
- Product schema on every product page, with price, availability, brand and GTIN. See our ecommerce schema guide.
- Structured product data in specification tables, not just images or paragraphs.
- Shipping and return policies as structured data and plain text.
- Consistent prices and stock across your site, feeds and marketplaces.
This is how to make products visible to AI — and it improves classic search and shopping results too. See also product page SEO.
Where agentic shopping happens
- ChatGPT shopping: product research and comparisons, with Instant Checkout for participating merchants.
- Google AI Mode shopping: built on Google’s product data and Shopping Graph. See how to optimise for AI Mode.
- Perplexity shopping: product answers with merchant links and purchase options in some markets.
We audit product data, feeds and schema for AI shopping agents.
Agentic commerce for retailers and B2B
- Agentic commerce for retailers: prioritise feed quality, reviews and fast, reliable fulfilment information.
- Agentic commerce for B2B: agents will increasingly handle reordering and supplier comparison — publish specifications, pricing tiers, lead times and procurement information clearly.
Agent-ready ecommerce checklist
- Complete, accurate product feeds and schema with GTINs.
- Specifications in structured tables.
- Prices and stock consistent everywhere.
- Shipping and returns in structured data.
- Genuine reviews on product pages.
- Checkout supported by your platform’s agentic options as they become available.
Agent-ready ecommerce is what AI agent optimization means for online stores. AI shopping optimization services focused on ecommerce AI visibility can audit this for you — see our AI shopping optimization and AI agent optimization services.
Agent-friendly checkout
An agent-friendly checkout is simple and predictable: guest checkout, clear shipping options and costs shown before payment, no surprise fees, and standard payment methods. Even before full agentic checkout reaches your platform, these choices help both human shoppers and AI agents for shopping.
Marketplaces and shopping assistants
Product comparison by AI also happens inside marketplaces: Amazon Rufus answers shopping questions using Amazon listings and reviews. Keep marketplace listings as complete as your own site, with price consistency across channels. Real-time inventory feeds stop agents recommending items you cannot ship.
Agentic commerce risks
Agentic commerce risks include fraud and agent payments (proving the user really authorised a purchase — the problem protocols such as AP2 address), returns from purchases the user did not fully review, and loss of direct customer relationships. Keep clear records, robust fraud checks and good post-purchase communication.
The future of ecommerce
Among AI shopping trends 2026, the clearest is that discovery is moving into assistants. The future of ecommerce will reward stores that are easy for both people and agents to understand: complete data, fair prices, reliable delivery and real reviews.
Agentic commerce: frequently asked questions
What is agentic commerce?
Shopping in which AI agents research, compare and sometimes buy products on a person’s behalf, based on a goal the person sets.
What is the Agentic Commerce Protocol?
A standard introduced by OpenAI with Stripe that lets merchants enable purchases directly inside ChatGPT through Instant Checkout.
What is AP2?
The Agent Payments Protocol, an open standard from Google for payments initiated by AI agents, designed to record that the user authorised the purchase.
How do AI shopping agents choose products?
They favour products with complete, accurate data, reliable pricing and stock, clear shipping and returns, strong reviews and consistent brand information.
How do I make my product data readable by AI?
Keep product feeds and schema complete with GTINs, put specifications in structured tables, keep prices and stock consistent, and publish shipping and return policies in structured form.
Do I need to change my ecommerce platform for agentic commerce?
Usually not. Most merchants will access agentic checkout through their platform and payment provider as support is added; product data quality is the part you control now.
What is Amazon Rufus?
Amazon’s AI shopping assistant, which answers product questions and compares items using Amazon listings and reviews.
Will AI agents replace online stores?
Unlikely. Agents still buy from stores; they change how shoppers discover and choose products, so data quality and trust matter more.
What are the risks of agentic commerce?
Fraud and authorisation questions, unwanted purchases and returns, and weaker direct relationships with customers.