On June 11, 2026, the US Ninth Circuit heard arguments on a question that touches every online store: can a user-authorized AI agent log in and act on a third-party platform on your behalf? The Amazon v. Perplexity hearing will produce the first federal precedent on the issue (Modern Retail). It is news because of an exploding number behind it: AI platforms are projected to drive about $20.57B in US retail ecommerce sales in 2026, nearly 4x 2025. Buying is shifting from "a human clicks your site" to "an agent completes checkout for them."
Market structure: over 12 months agentic commerce moved from concept to rails — OpenAI shipped instant checkout in ChatGPT and, with Stripe, the Agentic Commerce Protocol (ACP); Shopify and BigCommerce let merchants list their own agents. Why now? Automated and agent requests are mainstream — security firms observed bots at 57.4% of HTML web traffic in early June 2026, with agentic browsers (Perplexity Comet, OpenAI Atlas) the largest source. When machines browse your store more than humans do, "make the store machine-legible" goes from bonus to baseline.
Competitors and direction: Amazon sued Perplexity in January 2026 and on March 10 won a court order temporarily blocking Comet's shopping agent from its account areas. One camp guards its walled garden (ads, recommendations, data); the other (OpenAI/Stripe/Shopify) wants "agents check out directly" as an open protocol. The category is caught between protocol standardization and platform enclosure — and small merchants sit in the middle, both the "store" agents visit and the "seller" who might use agents to sell.
Below: what this means for your products being seen and bought by AI, what each role should do, the platform options and costs, and a DIY path to make your store agent-ready without an expensive SaaS subscription.
Event Details + Numbers
- Ninth Circuit oral arguments 2026/06/11 will set the first federal precedent on buyer-side AI agents accessing third-party platforms under user authorization.
- AI-driven US retail ecommerce sales ~$20.57B in 2026, nearly 4x 2025.
- Early June 2026, bots at 57.4% of HTML traffic, agentic browsers the largest source.
- Amazon sued Perplexity 2026/01; won a temporary injunction 2026/03/10.
- Protocol: OpenAI x Stripe Agentic Commerce Protocol and ChatGPT instant checkout are live.
Immediate Actions
- Brand owners/SMBs: confirm your products are findable and buyable in AI answers — do product pages have clear price, stock, shipping, and returns? Those are the fields agents read when comparing. Clean up product data before buying any agent SaaS.
- Marketers/SEO: add schema.org Product/Offer structured data (price, availability, reviews) and maintain a clean product feed (Google Merchant format) so agents and AI shopping UIs parse it correctly.
- Developers/agencies: evaluate instant-checkout/ACP integration and set an explicit "which agents we welcome, which crawlers we block" policy at robots/WAF — a technical decision, not default-open or default-block.
SaaS Comparison
| Platform/Option | What it does | Cost band | Best for |
|---|---|---|---|
| Shopify (built-in agent/Catalog) | Auto-lists products to AI shopping/agents | Monthly plan + tx fees | Already on Shopify, least effort |
| OpenAI Instant Checkout (ACP via Stripe) | Checkout inside ChatGPT | Stripe rate + platform cut | Want ChatGPT-driven sales |
| Google Merchant Center / AI Mode | Products in Google AI shopping | Free listing (ads extra) | Google-centric merchants |
| BigCommerce agent listing | Multichannel + agent-readable catalog | Monthly plan | B2B/multichannel |
What They Will Not Tell You
- "Agent-ready" can dilute your brand. When agents compare only price, shipping, and rating, your carefully designed brand story, upsells, and bundles get skipped. Agentic discovery pushes products toward pure spec-and-price competition — not great for premium-brand sellers.
- Neither protocol nor precedent is settled. The Ninth Circuit may limit or allow agent access; ACP keeps evolving. Heavy single-platform integration now may need redoing in six months. Spending on "clean, structured product data" — useful across every platform — is safer than betting on one agent channel.
SMB Alternative Without a SaaS Subscription
You do not need an expensive agentic-commerce SaaS first. Build an agent-ready base yourself: embed schema.org Product + Offer JSON-LD on product pages (price, currency, availability, return policy, reviews); export a standard product feed (Google Merchant / RSS) from your store or a script; define allowed/blocked agent UAs in robots.txt and at the server layer; ensure product pages are readable without login and have a fast LCP. This works across ChatGPT, Perplexity, and Google AI — the cost is mostly hours, not monthly fees.
FAQ
I am a small merchant — must I add instant checkout now?
No rush. Get structured data and your feed right first; that is the cross-platform base. Once you confirm agents actually drive orders, evaluate a specific checkout protocol — avoid early lock-in to a shifting channel.
Should I block AI agents or welcome them?
Separate them. Welcome shopping agents that bring orders and let them read your products; restrict content-scraping crawlers that do not refer traffic via robots/WAF. The point is a deliberate policy, not all-open or all-blocked.
Does structured data really get my products cited by AI?
It materially raises parse-ability and correct presentation. AI shopping UIs need explicit price/stock/shipping fields; without them, your products may be skipped even when seen. Necessary, not a guarantee.
My Take
The mainstream says "integrate agentic commerce or get left behind." My contrarian read: for most small merchants the best second-half-2026 move is not wiring up instant checkout but making product data machine-legible — precedent is unsettled and protocols shift, so the one thing that is never wasted is clean structured product data, useful in every AI channel. ScriptWalker takeaway: a clear productization opportunity — an "agent-ready audit" (product schema, feed, agent-access policy, page performance) makes a clearly-priced small project, far more pragmatic than hard-wiring a client to an agent platform still in court.