Square Puts Ordering Inside ChatGPT and Claude

Square added restaurant ordering to ChatGPT and Claude on July 1, 2026 — a launch that lets AI assistants discover Square-powered restaurants, read live menus, and place orders that drop straight into a seller’s existing point of sale. The framing is deliberately blunt: unlike the delivery marketplaces, Square charges no additional commission on these AI-channel orders.

That single design choice is the whole story. DoorDash, Uber Eats, and Grubhub take a percentage of every order — commonly 15% to 30% on delivery. Square’s ChatGPT and Claude channels take zero marketplace commission, so the only cost an operator pays is Square’s standard online processing fee. For a category that often nets single-digit margins, the difference is the difference between a channel that pays and a channel that bleeds.

This guide covers exactly what launched, how an order moves from an AI chat into a kitchen, the real fee comparison against the delivery apps (a table nobody else has built), the two vendor-cited numbers worth treating with care, and where this sits in the broader agentic commerce standards race. Every fact below is sourced to Square’s press release or the independent coverage that followed it.

On July 1, 2026, Square — the seller platform owned by Block — announced integrations that let ChatGPT and Claude discover Square-powered restaurants, surface their menus, and place orders directly inside the conversation. The initial rollout is US-only and limited to Food & Beverage sellers who already have an activated Square Online Ordering profile. Those eligible sellers are opted in automatically: no new APIs to build, no technical setup, no new contract to sign.

The strategic message from Block was that discovery is moving into AI assistants, and sellers need to be present there the way they once needed to be on the delivery apps. This is the local-SMB version of a shift that larger merchants have been preparing for through the discover-in-AI, buy-on-site pattern — except here Square owns both the discovery surface and the checkout rail.

The value of the integration is that it plugs into infrastructure a seller already runs, rather than bolting on a parallel system. When a customer asks an AI assistant to order from a nearby Square restaurant, the assistant reads live from that seller’s Square catalog — pricing, modifiers, and stock availability are dynamically mapped, so an agent cannot offer an item that is out of stock or quote a stale price.

Once the order is confirmed, it routes into the seller’s existing Square Point of Sale and Kitchen Display System exactly as an in-store or website order would — the kitchen does not need to watch a separate tablet. Square also tags the order source as an AI-channel order inside its reporting dashboard, so operators can track this channel’s volume and return separately from dine-in, web, and delivery.

Every piece of coverage described Square’s “no commission” claim qualitatively; none put the actual fee schedules side by side. So we built the table below from each provider’s current merchant pricing page, then applied every rate to the same representative $40 order. Marketplace commissions are verified at the primaries; Square’s figure is its standard online processing fee applied to the same order, since no separate “AI channel” fee tier exists.

The dollar column is where the claim becomes concrete. On a $40 delivery order, a DoorDash Premier or Uber Eats Premium seller hands over $12.00 in commission; a Square seller taking the same order through ChatGPT or Claude pays roughly $1.46 in processing on the Plus/Premium plan ($1.62 on the Free plan) and no commission at all — a gap of about $10.50 on that single order. The chart below shows the effective platform take rate on the same order.

The reason a 25%–30% commission is existential rather than annoying is the underlying margin. VentureBeat frames the math plainly: an independent restaurant clearing roughly 3%–9% net profit on a good day is effectively cooking at a loss when it hands a 25%–30% marketplace commission to a platform on a $40 order. A channel that removes the commission entirely does not just improve the economics of an order — for many operators it is the difference between an order worth taking and one that is not.

The honest caveat is delivery itself. Removing marketplace commission does not conjure a free courier. If an AI-generated order needs delivery, Square routes it through a white-label dispatch network that charges a flat courier fee rather than a percentage-of-basket commission — a structurally different, and for larger baskets usually cheaper, cost. But the specific fee figure is secondary-sourced, so treat it with care.

There is also a demand-side nuance worth keeping. PYMNTS’ own consumer research, published alongside the announcement, finds that people currently prefer AI as a collaborative shopping layer — for recommendations and comparisons — rather than a fully autonomous purchasing agent. That is a useful counterweight to the “agents buy for you” framing, and it maps to what we’ve seen in research on the consumer trust gap in agentic checkout. The near-term win for restaurants is discovery and a frictionless handoff, not a wave of fully autonomous ordering overnight.

Square’s release anchors the launch on two market numbers, and most coverage either repeated them without a source or dropped them silently. We think the more useful move is to name them plainly: both appear only in Square’s own press release, with no analyst firm or study cited. PYMNTS’ and DigitalCommerce360’s independent write-ups of the same announcement did not repeat either figure.

None of this undercuts the launch; the product is real and the fee advantage is verifiable at the primaries. It is a reminder that vendor announcements bundle real mechanics with directional market claims, and the two deserve different levels of trust. For hard numbers on the underlying market, we keep a maintained set of ecommerce statistics with named sources rather than vendor positioning.

The most strategically interesting part of the announcement is what surrounds it. Square is positioning this as one surface of a much larger agentic-commerce infrastructure play, and its own framing leans heavily on interoperability standards rather than a single walled channel. This is the local-SMB expression of the same standards race that runs through the UCP, ACP and AP2 standards reshaping how AI agents transact.

For an eligible Square restaurant, the near-term decision is simple: you are likely already opted in, so the work is measurement, not integration. For everyone else, the launch is a preview of where local commerce is heading and a prompt to get the fundamentals — a clean catalog, accurate stock, and a working online-ordering profile — in shape before AI discovery becomes table stakes.

For the ecommerce and restaurant brands we work with, the practical starting point is the same one we bring to any new channel: get the catalog and checkout genuinely agent-ready, then measure a live channel before betting on it. That is the core of our ecommerce growth engagements, and for teams standing up agent-facing commerce end to end it runs straight into our AI transformation work. The reference points from the assistant side are worth reading too — our guide to the ChatGPT shopping assistant covers what buyers actually experience on the other end of a flow like Square’s.