Square Turns On Restaurant Ordering Inside ChatGPT and Claude: Agentic Commerce Arrives
The Duckhub team builds AI-powered QR menu and online ordering software used by cafes, bars, and restaurants. We write practical guides based on what we see working across thousands of published menus.
On July 1, 2026, Square connected its US food & beverage sellers to ChatGPT and Claude, letting the AI assistants discover restaurants, read live menus, and place real orders — routed directly into Square Online Ordering and the restaurant’s POS, with no marketplace commission. Eligible sellers were automatically opted in, with no new APIs, setup, or contracts; the only cost is Square’s standard processing fee (roughly 2.9–3.3% + $0.30 per transaction). It is the first mass-market rollout of agentic ordering — AI completing the purchase, not just recommending the venue — across an entire restaurant platform.
What exactly launched
The integration works through the assistants’ app ecosystems: a guest asks ChatGPT or Claude to find food nearby or order from a specific restaurant, the assistant pulls the restaurant’s current Square menu, assembles the order conversationally (modifiers included), and submits it for pickup. The order lands in the restaurant’s existing Square ordering flow like any other digital ticket. Because enrollment was automatic for eligible US food & beverage sellers, hundreds of thousands of restaurants became AI-orderable overnight — most without knowing it.
The pricing structure is the strategic headline. Orders placed through the AI assistants carry no marketplace commission — a direct contrast with third-party delivery apps’ 15–30% per-order cut. Square is positioning AI surfaces as first-party channels: the platform monetizes payment processing it already owns, and the assistant becomes a storefront rather than a middleman.
Why this matters for restaurants
The demand-readiness gap this launch attacks is one of the widest in restaurant tech. Per the National Restaurant Association’s 2026 State of the Restaurant Industry, 26% of operators use AI tools somewhere in their business — but only 6% use AI to take customer orders, while roughly 6 in 10 millennial and Gen Z adults say they would order through an AI bot. Meanwhile, 45% of US consumers already use AI for local recommendations and 22% of diners have used AI to choose a restaurant — the discovery shift we track in restaurant marketing statistics. Agentic ordering closes the loop: the same assistant that picks the restaurant can now complete the sale.
Three practical consequences follow:
- Menu data just became sales infrastructure. An AI assistant can only sell what it can parse. Structured, live menu data — real item names, prices, modifiers — is the difference between being orderable in chat and being invisible. A PDF menu can’t take an order.
- The commission question resets. The first major agentic channels are launching commission-free, on first-party rails. Restaurants that route AI demand into owned ordering keep the margin that marketplaces would have taken — the same economics documented across our online ordering statistics.
- AI visibility now has a revenue line. With 83% of restaurant locations invisible in AI recommendations (Uberall), non-appearance stops being a branding problem and becomes lost orders.
The bigger agentic-commerce picture
Square’s launch is one move in an accelerating standards race: the company is co-developing agentic-commerce protocols and preparing further assistant integrations, while Google, OpenAI, and Anthropic each build the plumbing for AI-initiated transactions. For the restaurant industry — where digital ordering already runs through a decade-old stack of apps and aggregators — the open question is whether agentic channels stay commission-free once demand consolidates. The delivery-app era began with subsidized economics too.
For independents, the near-term playbook doesn’t require waiting on any platform: publish a structured, crawlable menu; keep hours, location, and items consistent everywhere; and connect an AI-readable ordering channel. Duckhub restaurants are already reachable this way — through Duckhub’s ChatGPT app and MCP integration, AI assistants can browse a venue’s live menu and support ordering at 0% commission; as of this writing, no other dedicated QR-menu platform offers an equivalent. Setting up takes about 5 minutes on the 30-day free trial.
The one-paragraph summary: On July 1, 2026, Square made US restaurants orderable inside ChatGPT and Claude — automatically, commission-free, with standard processing as the only fee — at a moment when only 6% of restaurants take AI orders but a majority of younger diners say they want to. Agentic commerce for restaurants is no longer a forecast; it’s a settings page.
Reported from Square’s press announcement and NRA 2026 survey data, verified July 2026. Related data hubs: AI in restaurants statistics · online ordering statistics.