News
Square Put Sellers Inside ChatGPT and Claude. The CFO Read on Agentic Commerce.
Square put sellers inside ChatGPT and Claude, so AI assistants can discover them and place orders during a conversation, per Digital Commerce 360. US food and beverage sellers are live first, with no extra commission for now. This is a new discovery and checkout channel with unknown economics. The move is to get your product data present and measurable on these surfaces, then model the channel before its take rate arrives.
Key Takeaways
- Square launched a ChatGPT app and Claude plugin for agentic commerce: AI assistants can now discover eligible sellers and place orders inside a conversation. US food and beverage sellers on Square Online Ordering are live first.
- The onboarding is frictionless by design. Eligible sellers are auto opted-in with no setup, API work or fees, and Square is charging no additional marketplace commission on these orders for now. Read that as adoption-seeding, not a permanent price.
- This is a genuinely new channel: discovery and checkout happen inside an AI assistant, not on your site or a marketplace. The economics are unknown, which is exactly why you model the framework now, before a take rate lands.
- Agentic commerce shifts discovery from search and paid to being present in an AI agent's consideration set, and it makes attribution harder because the agent mediates the journey. Feed readiness and measurement become the new blocking and tackling.
- The move today is to position, not to bet big. Make sure your product and catalog data is structured and present on these surfaces, decide how you will measure the channel, and keep the investment sized to a channel that is still forming.
Square just made it possible for an AI assistant to find one of its sellers and place an order without the customer ever visiting a website. Ask ChatGPT or Claude to order dinner, and an eligible Square merchant can now be discovered and transacted with inside the conversation. It is easy to file this under interesting AI news. For an operator, it is the early shape of a new channel, and new channels are worth understanding before their economics harden.
We think about a surface like this the way we think about any channel: what does it do to discovery, to attribution and to contribution margin. It is an extension of the shift we track in AI-driven discovery and LLM discoverability. Here is the CFO read on agentic commerce.
What happened
Digital Commerce 360 reported that on July 1, 2026, Square launched a ChatGPT app and a Claude plugin for agentic commerce, letting eligible sellers be discovered and accept orders inside AI conversations. The first live merchants are US food and beverage sellers with active Square Online Ordering profiles, and orders route straight into their existing Square POS, Kitchen Display and reporting.
The onboarding is deliberately frictionless. Eligible sellers are automatically opted in, with no additional setup, API work or fees, and Square says it charges no extra marketplace commission on these orders. Square, which serves more than 4.5 million sellers, framed this as infrastructure it handles so merchants do not have to. Retail, services and appointment-based commerce are described as coming soon.
Square is also positioning across the emerging standards: it has joined the AAIF Agentic Commerce Working Group and the W3C Web Payments Working Group, and it is working with the Universal Commerce Protocol effort led by Google and Shopify. Amazon's Alexa+ voice commerce and Order by Cash App are part of the same broader push.
| Square agentic commerce, July 2026 | Detail |
|---|---|
| Surfaces | ChatGPT app, Claude plugin |
| First live | US food and beverage on Square Online Ordering |
| Fees | No setup, API work or added commission (for now) |
| Order flow | Into existing Square POS and reporting |
| Seller base | 4.5M+ sellers |
| Standards | AAIF, W3C Web Payments, UCP (Google, Shopify) |
Source: Digital Commerce 360, Square and VentureBeat, July 2026.
Returns are quietly eating your margin. See by how much.
Get our Real Cost of Returns calculator: plug in your numbers, see the true hit per return.
Check your inbox. We'll send the Real Cost of Returns calculator shortly.
A new channel, and why the free price is a signal
Start with the economics, because they are the most misread part. The headline that Square charges no extra commission is real and attractive, but it should be read as a stage, not a permanent feature. New channels routinely launch free to seed supply and demand, then introduce a take rate once volume and habit exist. We have watched this pattern in every channel that came before, and there is little reason agentic commerce will be different once the surfaces have the traffic.
That does not mean avoid it. It means use the free window well. The value of a no-fee launch is that you can build presence and, more importantly, the muscle to measure the channel, before its cost structure is set. When you already sell through several channels, you know the discipline: the contribution margin by channel view is what tells you whether a channel is worth its cost. Agentic commerce has no take rate to plug into that model yet. The job today is to be ready to judge it the moment it does.
Discovery moves inside the agent
The bigger shift is where discovery happens. For years, being found meant ranking in search, showing up in social, and buying paid placement. Agentic commerce introduces a different surface entirely: the consideration set an AI assistant draws on when a customer asks it to find or order something. If your products and data are present and legible to those agents, you can be surfaced without paying for the click. If you are absent, no amount of traditional SEO puts you in that conversation.
This is the same movement we have been tracking in how AI changes traffic and discovery, the logic behind our work on modeling AI traffic for a brand and on conversion by traffic source. The practical implication is that feed and data readiness becomes a growth lever, not an IT chore. Structured, accurate, current product data is what makes you discoverable to an agent. The brands that treat that as strategic will be present when a customer's assistant goes looking. The ones that do not will simply be invisible in an increasingly common way of buying.
Attribution gets harder, so decide how you will measure it
The cost of an agent mediating the purchase is that you lose sight of the journey. When ChatGPT or Claude handles discovery and checkout, the order can land in your system without the funnel data you are used to: which alternatives the customer saw, what they searched, what tipped the decision. That is a real problem, because CAC, conversion and channel contribution all depend on being able to see the path.
So the move is not to wait for perfect attribution, it is to decide up front how you will measure agentic orders even imperfectly. Tag them, segment them, and build the ability to compare their contribution and cost against your other channels over time. This is exactly the measurement discipline we apply to any new source of traffic and cost, in the same spirit as our customer acquisition cost by channel work. You cannot manage a channel you cannot see, and agentic commerce will stay a black box unless you deliberately instrument it.
What to watch next
- The arrival of a take rate. The channel is free now. Watch for the introduction of commissions or fees, and be ready to re-run channel contribution the moment they land. A channel that is profitable at zero fee may not be at ten percent.
- Your data and feed readiness. Agents surface what they can read. Keep your catalog, pricing, availability and attributes structured, accurate and present on the platforms building agentic commerce. This is the cheapest, highest-leverage thing you can do today.
- Measurement of agentic orders. Put even rough tagging in place now so that agentic volume does not grow into a blind spot. When the channel matters, you will want a contribution and cost history, not a standing start.
The operator takeaway
Square putting sellers inside ChatGPT and Claude is a small, category-limited step, but it points at something larger: buying is starting to happen inside AI assistants, and that is a new channel with its own discovery rules, its own attribution problems, and economics that are not settled yet. The mistake would be to dismiss it as hype, or to over-invest in it as if it were proven. Neither is right this early.
The measured response is to position without betting the business. Get your product data clean and present on these surfaces so you are discoverable, decide how you will measure agentic orders before volume arrives, and keep the investment sized to a channel that is still forming. Then let the data tell you when to lean in. That is how you take the option on a new channel cheaply and scale it only when it earns it. If you want help sizing that decision for your brand, our team does exactly this.
Frequently Asked Questions
what did square announce with chatgpt and claude?
On July 1, 2026, Square launched a ChatGPT app and a Claude plugin that let its sellers be discovered and accept orders inside AI-assistant conversations. The first live merchants are US food and beverage sellers with active Square Online Ordering profiles, and orders flow straight into their existing Square POS, Kitchen Display and reporting systems. Eligible sellers are automatically opted in, with no additional setup, API work or fees required, and Square says it does not charge extra marketplace commissions on these orders. Retail, services and appointment-based commerce are described as coming soon.
what is agentic commerce?
Agentic commerce is when an AI assistant, like ChatGPT or Claude, does the shopping on the customer's behalf: finding options, comparing them and completing the purchase inside the conversation, rather than sending the person to a website to click through a checkout. Instead of a customer browsing your store, they ask an agent for something, and the agent surfaces and transacts with whichever merchants are available to it. It is a new layer between the shopper and the seller. For merchants, it turns discovery and checkout into something that happens inside someone else's interface.
does square charge a commission on ai-driven orders?
Not right now. Square says eligible sellers are auto opted-in with no setup, API work or fees, and that it is not charging additional marketplace commissions on orders placed through the ChatGPT and Claude integrations. That is attractive, but treat it as an adoption-seeding price rather than a permanent one. New channels frequently launch free to build supply and demand, then introduce a take rate once they have volume and habit. The smart move is to enjoy the free window while building the muscle to measure the channel, so that if and when a fee arrives you can judge the channel on its true economics.
what does agentic commerce mean for a dtc brand's discovery?
It adds a new place you need to be found. Today, discovery runs through search, social and paid, and you optimize for those. Agentic commerce introduces a different surface: the consideration set an AI assistant draws on when a customer asks it to find or order something. If your products and data are present and legible to those agents, you can be surfaced without paying to acquire the click. If you are absent, you are invisible in that conversation regardless of how good your SEO is. It is an extension of the same shift we track in AI-driven discovery, where being readable by models becomes its own channel.
how does agentic checkout change attribution and measurement?
It makes the journey harder to see. When an agent mediates discovery and checkout, you lose visibility into the steps a customer took, the alternatives they were shown and the context of the purchase. The order can arrive in your system without the usual funnel data attached. That matters because you cannot manage what you cannot measure: CAC, conversion and channel contribution all get murkier when the agent owns the interaction. Before you lean into the channel, decide how you will tag and measure agentic orders, even imperfectly, so you can eventually compare their contribution and cost against your other channels.
should i invest in agentic commerce now or wait?
Position now, but do not bet big yet. The channel is real and moving quickly, with Square, Amazon and protocol groups like the Universal Commerce Protocol all building toward it, so being completely absent is a risk. At the same time, it is early, largely limited to specific categories, and its economics are unproven. The right posture is to make sure your product and catalog data is structured and present on these surfaces, put light measurement in place, and keep the spend sized to a forming channel. Get the option to participate cheaply, and scale only when the data shows real, profitable volume.
how do i get my brand ready for ai agent discovery?
Start with your data, because agents can only surface what they can read. Make sure your product catalog, pricing, availability and key attributes are structured, accurate and accessible on the platforms building agentic commerce, whether that is your commerce platform, a payments provider like Square, or a marketplace. Keep information like hours, inventory and descriptions current, since agents sync and act on it in real time. Then decide how you will identify and measure orders that come through agentic surfaces. The brands that win the next channel are usually the ones whose data was clean and present before it mattered, not the ones who scrambled after.