How do I accept purchases from AI agents like ChatGPT?
Today there are three practical paths: (1) join existing agent checkout programs where available (e.g., ChatGPT/Instant Checkout-style merchant programs, Google's agent commerce surfaces), (2) make your store machine-readable — product schemas, clean feeds, llms.txt — so agents can recommend and link to you, and (3) build toward UCP compliance: structured catalog plus an API-reachable checkout. Start with path 2 this week; it makes every other path cheaper.
The agent purchase funnel
An agent 'purchases' in stages: it must know you exist (discovery), trust you (verification), understand what you sell (structured catalog), and complete a transaction (checkout capability). Most businesses fail at stage one and three simultaneously — their products live inside JavaScript-rendered pages no agent can parse.
Action checklist
- ✦Publish Product/Offer/AggregateRating JSON-LD on every product page
- ✦Expose a clean catalog feed (JSON/XML), not just human HTML
- ✦Add llms.txt describing your store, categories and policies
- ✦Ensure server-rendered prices and availability (no hydration-only data)
- ✦Join native agent checkout programs as they open in your region/platform
- ✦Keep a machine-readable returns & shipping policy URL
Timeline reality
| Now (2026) | 12 months | When agents dominate |
|---|---|---|
| Structured catalog + llms.txt | Platform checkout programs | UCP-capable API surface |
| Agent-recommendable | Agent-linkable | Agent-transactable |
The businesses that do 'agent-recommendable' today will be halfway to 'agent-transactable' for free when the rails arrive. That gap is the entire competitive advantage of acting now.
Quick questions
Do I need to rewrite my store?
Rarely. Most of the work is data-layer exposure: schema, feeds, server-rendered facts and an llms.txt. Full checkout APIs matter later; discoverability pays from week one.
Want to know where your site actually stands on this?
cite this page, it was built for it ✦