Agentic Commerce: How AI Agents Shop for Customers in 2026

Your customer just bought — and never clicked your website

Here's what a purchase looks like in 2026: A customer asks ChatGPT to "find me the best running shoes under €120 with fast shipping." The AI searches structured product feeds, surfaces three options with availability and price, presents a checkout button, and the transaction is complete. No Google search. No product page visit. No cart abandonment. Your store either appeared in that recommendation — or it didn't.

This is agentic commerce: AI agents acting as autonomous shopping intermediaries between customers and retailers. ChatGPT's Instant Checkout has been live since September 2025, processing purchases for 900 million weekly users. Google launched its own competing standard in January 2026, backed by Walmart, Target, and Shopify. The funnel has fundamentally changed.

Two protocols now control AI shopping

AI agents needed reliable, standardized APIs to replace their earlier approach of scraping retailer websites and breaking on checkout forms. Two competing protocols have emerged as the infrastructure layer for agentic commerce — and brands need to understand both.

Factor	ACP (Agentic Commerce Protocol)	UCP (Universal Commerce Protocol)
Owner	OpenAI + Stripe	Google + Coalition
Live since	September 2025	Announced January 2026
Where it operates	ChatGPT (web + mobile)	Google AI Mode, Gemini
Key partners	Stripe-enabled merchants	Walmart, Target, Shopify, 20+ others
Best for	Conversational product discovery	High-intent search queries
Required data	Structured product feed, real-time inventory	Schema.org markup, product API

This is not a winner-takes-all format war. Both protocols will coexist, serving different user contexts — the same way brands run Google Ads and Meta Ads in parallel. Brands that support only one will miss half the agentic channel.

What AI agents need to recommend your store

AI shopping agents don't browse. They call APIs and parse structured data. If your product information isn't machine-readable, the agent skips you entirely — not because your product is worse, but because it can't read what you're selling. Four things determine whether an AI agent can surface your products:

• Structured product data: Schema.org Product markup with accurate price, availability, description, and images. Missing or outdated fields cause agents to deprioritize or skip your listings.
• Real-time inventory access: AI agents won't recommend an out-of-stock product. Your feed or API must reflect live inventory — static sitemaps are insufficient.
• AI crawler access: GPTBot, PerplexityBot, and ClaudeBot must be explicitly allowed in your robots.txt. Many stores block them accidentally through CDN or firewall rules.
• Direct-answer product descriptions: Content written to answer "which [product] should I buy and why?" performs dramatically better than keyword-optimized copy.

The revenue case for acting now

Morgan Stanley projects that nearly half of online shoppers will use AI agents for purchases by 2030, accounting for approximately 25% of total online spending. That's not a future scenario — the migration is already underway. AI-referred shoppers convert at 38% higher rates than traditional organic search visitors, because they arrive with specific, pre-qualified intent. The agent already filtered and recommended. By the time the purchase happens, the decision is made.

The compounding risk for stores that wait: AI agents develop citation patterns. A model that recommends your competitor 500 times builds a reinforcement loop that's difficult to displace. Early visibility now creates a structural advantage in 12 months. Brands that start AEO and agentic commerce preparation in 2026 will have compounding data, agent familiarity, and conversion history that latecomers cannot easily replicate.

How to make your store agent-ready this quarter

Getting agentic-commerce-ready doesn't require a platform rebuild. The changes are mostly data and configuration — but they need to be done correctly, because AI agents either read your data cleanly or not at all.

• Audit your robots.txt: explicitly allow GPTBot, PerplexityBot, ClaudeBot, and Google-Extended
• Implement Schema.org Product markup on all product pages (price, availability, description, brand, image)
• Create or update your llms.txt file with a clear map to your catalog, category pages, and FAQs
• Build or expose a product feed API that returns real-time inventory and pricing
• Rewrite product descriptions to answer the question "why should I buy this over alternatives?"
• Add Organization and LocalBusiness schema to establish trust signals for AI recommender systems
• Test your store by directly asking ChatGPT and Perplexity to recommend products in your category

Frequently Asked Questions

What is agentic commerce?

Agentic commerce is when an AI shopping agent — such as ChatGPT, Gemini, or Perplexity — searches, compares, and purchases products autonomously on behalf of a customer, without the customer visiting a retailer's website directly. The agent reads machine-readable product data, selects the best match for the user's goal, and completes checkout through a commerce protocol like ACP (OpenAI + Stripe) or UCP (Google + Coalition).

What is the difference between ACP and UCP?

ACP (Agentic Commerce Protocol) is OpenAI and Stripe's standard, live since September 2025, powering ChatGPT Instant Checkout for 900 million weekly users. UCP (Universal Commerce Protocol) is Google's competing standard from January 2026, backed by Walmart, Target, and Shopify, operating inside Google AI Mode and Gemini. Both are API-first protocols that let AI agents access product catalogs and complete transactions. Brands need to support both to cover the full agentic commerce landscape.

How do I make my online store visible to AI shopping agents?

Your store needs: (1) Schema.org Product markup on all product pages; (2) an llms.txt file listing key pages for AI crawlers; (3) robots.txt that explicitly allows GPTBot, PerplexityBot, and ClaudeBot; (4) a real-time product feed or API; and (5) product descriptions written as direct answers to shopping questions. Without these, AI agents cannot read or recommend your products regardless of how strong your SEO is.

Will agentic commerce replace regular website traffic?

Not entirely, but the share of purchases initiated through AI agents is growing fast. Morgan Stanley projects that nearly half of online shoppers will use AI agents by 2030, accounting for approximately 25% of total online spending. Stores that are not agent-ready will lose a disproportionate share of high-intent purchases to competitors who are prepared.

How much revenue will agentic commerce generate by 2030?

McKinsey projects that agentic commerce will drive $3 to $5 trillion in global commerce by 2030, including both direct AI-agent purchases and AI-assisted discovery that ends in a traditional checkout. The channel is expected to grow faster than mobile commerce grew between 2012 and 2018.