A week ago, a hungry person near your restaurant had a few ways to find you. Google. A delivery app. Driving past your sign. As of July 1, there’s a new one, and it’s the one nobody’s watching: they open ChatGPT or Claude and type “find me good tacos near me that I can order right now.”
If your restaurant is set up right, the AI names you, shows your menu, and takes the order — routed straight to your kitchen, with no 30% delivery commission skimmed off the top. If you’re not set up right, it names the place down the street and you never even know you were in the running.
That’s the shift. AI didn’t just become a thing your customers talk to. It became your new front door. And most restaurants are about to walk right past it because nobody told them the door exists.
Let me walk you through what actually happened, how to check if you’re already live (a lot of you are and don’t know it), and the handful of things that decide whether the AI picks you.
What Just Changed
On July 1, 2026, Square rolled out a ChatGPT app and a Claude plugin that plug your restaurant directly into those AI assistants. Here’s the plain version of how it works:
A customer asks ChatGPT something like “where can I get a burger near me and order it here.” The AI looks at Square restaurants nearby, matches your menu and hours to what they asked, shows them your food, builds the order as they chat, and sends it straight into your existing Square system — the same Point of Sale and kitchen screen you already use. They pay through Square. The ticket prints in your kitchen like any other online order.
You don’t build a bot. You don’t sign up for an OpenAI account. You don’t manage a new listing. If you’re an eligible Square restaurant, Square switched it on for you.
And here’s the number that matters: Square charges no marketplace commission on these orders. Just its normal card processing — roughly 2.9% to 3.3% plus about 30 cents, the same rate you already pay. Compare that to the 15% to 30% that DoorDash and Uber Eats take, and you can see why restaurant people are paying attention. One post going around from the Square team put it bluntly: the “aggregator tax” is the first thing that dies.
This isn’t a one-off, either. Google, Amazon (through its Alexa+ assistant), and a handful of startups are all wiring restaurants into AI ordering. Analysts are throwing around numbers like $385 billion in “AI does the shopping” transactions by the early 2030s. Restaurants are one slice of that. Square just happened to make it real for independent spots first — with no setup and no code.
Are You Already On It? Check in 10 Minutes
Because Square auto-enabled eligible sellers, a lot of restaurant owners are already live and have no idea. So before anything else, find out where you stand.
You’re likely eligible if:
- You’re a US Food & Beverage seller on Square
- You use Square Online Ordering or Square for Restaurants (you have a real menu set up online)
- Your account is in good standing
To check: open ChatGPT or Claude and ask it to find and order from a restaurant like yours in your area. See if you come up. Then poke around your Square Online Ordering dashboard for the AI ordering or discovery setting. If you’re eligible, it’s on by default — and yes, there’s an opt-out if you decide you don’t want AI orders (though I’d think hard before turning down commission-free tickets).
If you’re not on Square at all, that’s a bigger conversation. But the trend is the point here: AI assistants are becoming a place people order food. Square is first. It won’t be last.
The Fee Math (Why This Actually Matters)
Restaurant margins are thin. Brutally thin. So let’s do the math on a single order, because this is where it gets real.
Say a customer orders $40 of food.
- Through a delivery marketplace at 25% commission: the platform takes $10. You keep $30 to cover the food, the labor, the rent, and — if you’re lucky — a little profit. On a lot of orders, that $10 cut turns your profit into roughly zero.
- Through Square’s ChatGPT/Claude ordering: you pay about $1.50 in processing. You keep roughly $38.50.
Same order. Same food out the door. An extra eight-and-a-half bucks in your pocket instead of a delivery giant’s. Run that across a few hundred orders a month and it’s not a rounding error — it’s a raise for your business.
The frustration this taps into is real and loud. One viral post showed a delivery order ringing up at $37.17 that cost $20.96 to walk in and buy — an 80% markup, most of it fees. Customers hate it. You hate it. This is the first tool that lets a normal, non-technical owner route around it without becoming a tech company.
How AI Decides Which Restaurant to Recommend
Here’s the catch, and it’s a big one. Getting the AI to take an order is easy — Square handles that. Getting the AI to pick you in the first place is the actual work. And most restaurants are failing it.
A May 2026 study from Uberall found that 83% of restaurants are completely invisible in AI recommendations. When someone asks ChatGPT “where’s good pizza near me tonight,” only about 17% of places ever show up in the answer — even though 86% of them have a Google listing. There’s no page two in an AI answer. You’re either in the short list it reads out, or you don’t exist.
So what gets you into that short list? A few things, and none of them require a marketing degree.
Your reviews clear the bar. This one surprised me. The AI assistants each seem to have a rough rating floor before they’ll recommend a place. ChatGPT leans toward restaurants averaging around 4.3 stars or higher. Perplexity, roughly 4.1+. Google’s Gemini is more forgiving, around 3.9+. A 4.0 restaurant can rank fine on Google and still fall below the line the AI uses. Recent reviews matter more than a pile of old ones, so a steady trickle of fresh, positive reviews beats a big stale number.
Your info is consistent everywhere. Your name, address, hours, and phone need to match across Google, Maps, Yelp, and your own site. When the AI finds three different phone numbers for you, it gets confused about which “you” is real — and confused means skipped.
Your menu is written the way people ask. This is the sneaky one. Diners don’t ask for “our signature protein bowl.” They ask for “gluten-free lunch near me” or “best birria tacos.” If your menu names and descriptions use the words real people search — actual dish names, real descriptions, dietary tags — the AI can match you. Generic, chef-flowery menu copy is invisible. Named dishes are findable.
A 20-minute menu tune-up
Grab your menu and a free ChatGPT window. Try a prompt like:
“Here’s my restaurant menu. Rewrite each item name and description the way a hungry customer would actually search for it — plain language, include the dish type and any dietary tags like gluten-free or vegan. Keep it appetizing but searchable.”
Paste your items in, and you’ll get menu copy built for how people (and the AI reading on their behalf) actually look for food. Do the same for your hours and your Google listing. Twenty minutes, and you’ve moved yourself toward that 17% that shows up instead of the 83% that doesn’t.
What This Means for You
Depends on your spot. Find yourself here.
If you’re a busy independent — one location, tight margins: this is the closest thing to found money you’ll see this year. Check if you’re live, tune your menu wording, and start nudging happy regulars for reviews. The goal isn’t to master AI. It’s to clear the 4.3-star bar and keep your info clean so the assistant names you when a neighbor asks.
If you run a café or quick-service spot: speed and repeat orders are your game, and AI ordering rewards both — no app tax, straight to your kitchen. Make sure your most-searched items (the breakfast sandwich, the oat-milk latte) are named plainly so the AI can match a craving to your counter.
If you’re a food truck or pop-up: your hours and location change, which is exactly what confuses AI. Keeping your Square profile and Google hours dead accurate is your single highest-leverage move. An AI that knows you’re open right now is an AI that can send someone hungry your way.
If you barely touch tech: you don’t have to become a “tech restaurant.” The whole point of Square doing this is that you don’t build anything. Do three things — confirm you’re live, fix your menu wording, ask for reviews — and you’ve done 90% of what matters. A short course can walk you through each one if you’d rather be shown than figure it out.
What This Can’t Fix
I’m not going to oversell it. Here’s the honest side.
The customers aren’t all there yet. Let’s be real — most people still open a delivery app, not ChatGPT, when they’re hungry. Ordering food by chatting with an AI is early. Right now the tech crowd is doing it and normal folks mostly aren’t. You’re planting a flag before the crowd arrives, which is smart, but don’t expect a flood of AI orders next week. This is a “be ready” play, not a “get rich Tuesday” play.
The AI flattens discovery. When someone browses a delivery app, they scroll, they impulse-add a milkshake, they discover you by accident. An AI that fetches exactly what someone asked skips all of that — one company found shoppers convert far less when an assistant grabs one item instead of letting them wander. So AI ordering is efficient, but it can strip out the browsing and upselling that used to help you. You win by being the answer, not by being stumbled upon.
It can quietly favor the popular. A system that recommends the highest-rated, most-reviewed places can tilt toward chains and established spots over the new place with 12 reviews. That’s not a reason to sit out — it’s the reason to start building reviews and clean data now, so you’re not invisible when it matters.
It’s US-only, for now. If you’re outside the States, the Square version isn’t live for you yet. The playbook — good reviews, clean info, searchable menu — still travels. The ordering rail will follow.
The Bottom Line
AI just became a place people find and order dinner, and for the first time it doesn’t cost you a third of the ticket. The restaurants that win the next couple of years won’t be the ones with the fanciest tech. They’ll be the ones that showed up early, kept their reviews above the line, and wrote their menu the way hungry people actually talk.
You can do the whole starter version this week: check if you’re live, run the 20-minute menu tune-up, and start asking happy customers for a review. That’s it. That’s the front door, propped open.
If you’d rather be walked through it step by step — getting found, getting recommended, and taking those commission-free orders — our AI for Restaurants course lays out the exact playbook for non-technical owners. First lessons are free, no signup.
The diner asking ChatGPT for “good food near me tonight” is real, and there are more of them every week. Make sure the answer is you.
Sources
- Square Introduces New ChatGPT and Claude Integrations — Square Press (July 1, 2026)
- Restaurants can now accept orders placed directly from ChatGPT and Claude — VentureBeat (July 1, 2026)
- Square launches agentic commerce integrations with ChatGPT, Claude — Digital Commerce 360 (July 1, 2026)
- 83% of Restaurants Are Invisible in AI Search — Uberall report (May 2026)
- Square puts sellers inside ChatGPT and Claude — the CFO read on agentic commerce — Eightx (July 2026)
