AI Assistants Are Now Taking Orders and Bookings — Here's How a UK Restaurant or Salon Gets Included
Square switched on ordering inside ChatGPT and Claude on 1 July 2026 with no marketplace commission — here's the free groundwork a UK restaurant or salon should do now to be the business an AI assistant recommends.
On 1 July 2026, Square switched on restaurant ordering directly inside ChatGPT and Claude. A customer can ask an assistant to find food nearby and place the order without ever opening a website or a delivery app, and Square is charging no marketplace commission on those orders. Compare that to the 15–30% a delivery marketplace typically takes and you can see why UK operators are paying attention, even though the first rollout is not a UK-wide switch-on.
The honest position for a UK venue today: you cannot force your way into this yet, but you can make yourself readable. Almost every step below is free, takes an afternoon, and helps your ordinary Google visibility too. That's the whole argument for doing it now rather than waiting.
What actually happened on 1 July 2026?
Square, the till and payments company, built ordering into ChatGPT and Claude so that eligible restaurants using Square can receive orders placed inside a conversation with an AI assistant. Square is not charging a marketplace commission on those orders — the restaurant still pays normal card processing, but not the double-digit cut a delivery platform takes.
That's the significant bit. The delivery app model works because it owns discovery: the customer opens Deliveroo, not your website. If discovery moves into an assistant, and the assistant hands the order straight to your till, the middleman's cut becomes harder to justify.
What does "agentic ordering" mean for a small venue?
Agentic ordering means the customer states an outcome and software does the steps. Instead of "show me Thai restaurants", it's "order me a green curry, medium spice, for collection at 7pm from somewhere well-reviewed within a mile" — and the assistant picks the venue, builds the basket and pays.
The practical consequence: you are being shortlisted by a machine before a human sees your name. There is no shopfront photo doing the persuading, no carousel of images. The assistant reads text about you, decides you fit, and presents one to three options. If your details are thin, wrong or trapped in a format it can't parse, you are not on that shortlist and you will never know you were considered.
Is AI ordering live in the UK yet?
Not as a general UK switch-on. Square's ChatGPT and Claude ordering went live on 1 July 2026 and rollouts of this kind typically start in the US and widen over the following months, with payments, tax and delivery rules to work through per market. Treat a UK date as unannounced rather than imminent.
What is already live in the UK is the recommendation half. People here are asking ChatGPT, Claude, Gemini and Google's AI answers for "best Sunday roast near Didsbury" or "balayage specialist in Bristol who does Saturdays" right now, and getting a shortlist of three named businesses. That happens today, without any till integration, and it is the part you can influence.
How does an AI assistant decide which business to recommend?
Assistants assemble an answer from indexed web pages, business listings and review sites. In practice, the details that decide whether you make the cut are:
| What it reads | What it wants to find | Common failure |
|---|---|---|
| Business profile | Name, address, phone, category, all identical everywhere | Old phone number on an unclaimed listing |
| Opening hours | Current hours, including bank holidays | "See Facebook for hours" |
| Menu or service list | Dish or service names, descriptions, prices, in text | PDF, image, or Instagram-only |
| Service area | Named towns, postcode districts, radius | "We cover the local area" |
| Reviews | Recent, and mentioning specific dishes or treatments | 11 reviews from 2021 |
| Dietary and specialisms | Gluten-free, halal, vegan, kids, dog-friendly, step-free | Nothing stated, so nothing matched |
Notice how many of those are filters. When someone asks for "gluten-free pizza, open now, collection in Leeds", the assistant needs four facts to include you. Miss any one and you drop out — not because you're worse, but because you're unverifiable.
The strongest single signal you control is specificity. "Family-run Neapolitan pizzeria on Chapel Allerton high street, gluten-free bases £2, open until 10pm Thursday to Sunday, collection only" gives an assistant six things to match against. "Great food, friendly service, come and see us" gives it nothing.
Why does a PDF or image menu make you invisible?
Because the words in it aren't words as far as most crawlers are concerned. A PDF menu is a separate file that often isn't indexed properly, and a photographed menu or a chalkboard picture is just pixels. If your carbonara only exists inside a JPEG, an assistant asked for "carbonara near me" cannot see that you sell it.
This is the single most common own goal in UK hospitality. The menu is the most valuable text your business owns — it's every search term a hungry person will ever type — and it's sitting in a file nobody can read.
The fix is a menu page in ordinary HTML: headings for courses, each dish as text with a short description and a price. Same logic for salons: a services page listing "cut and finish, 45 minutes, £38" beats "prices on request" every time. Keep the PDF for printing if you like, but the web version must be text.
While you're at it, adding Menu or Restaurant schema markup (or LocalBusiness and Service for salons) makes the same facts machine-readable in a structured way. Most decent website tools do this without you touching code — SiteGenie generates it automatically for the pages it builds — but the plain-text menu matters far more than the markup. Get that right first.
What's the free groundwork to do this month?
Work through these in order. None of it costs money.
- Claim and complete your Google Business Profile. Category, hours, holiday hours, attributes, service area, phone, website. Add your menu or service list inside the profile as well as on your site.
- Publish your menu or services as text on your own site. Names, one-line descriptions, prices. Update it when prices change rather than uploading a new PDF.
- Make hours unambiguous. State them on the site in text, and match them exactly to Google, Facebook and any booking platform. Conflicting hours make an assistant cautious about recommending you at all.
- Name your area. "Collection from our Stoke Newington shop, delivery to N16 and N4" is usable. "Local deliveries" is not.
- Write the specialisms down. Halal, vegan options, gluten-free, high chairs, step-free access, late opening, dog-friendly, walk-ins welcome, kids' cuts, bridal hair. Each one is a query you can win.
- Ask for reviews that describe things. A review saying "the lamb shawarma and the baklava were superb" is worth far more to an assistant than five stars with no text. Ask customers to mention what they had.
- Check your basics on the aggregators. TripAdvisor, Yell, Treatwell, OpenTable, Bing Places. You don't need to be active on all of them, but wrong details on any of them muddy the picture.
- Make sure your own booking or ordering route works on a phone. If an assistant sends someone to your site, a broken form wastes the referral.
That's a couple of hours of admin and it improves your Google results, your Maps ranking and your conversion rate regardless of what happens with AI ordering.
What about salons and appointment businesses?
Same mechanics, different verb. People are already asking assistants for "a barber in Cardiff open Sunday" or "gel removal near Leeds station this afternoon", and the assistant needs named services, durations, prices, staff availability and a bookable link to answer well.
The specific trap for salons is the booking widget. If availability lives inside an embedded calendar that loads after the page does, an assistant can't read it. So publish the facts in text next to the widget: which services you offer, how long each takes, what it costs, which days you're open late, whether you take walk-ins. Then let the widget handle the actual booking.
If you use Fresha, Treatwell, Booksy or Square Appointments, keep those listings as accurate as your own site. Booking platforms are exactly the sort of integration partners that get wired into assistants next.
Should I drop the delivery apps?
No, not on the basis of one launch in another market. Deliveroo, Just Eat and Uber Eats still bring volume you can't replicate overnight, and a commission you resent is still better than an empty kitchen.
The sensible read is that discovery is starting to split. Some customers will keep opening a delivery app; some will ask an assistant; some will just search. Your job is to be findable in all three, and to nudge repeat customers towards the channel that doesn't take 15–30% — a direct ordering page, a collection option, a loyalty text. That has been good advice for years. AI ordering just raises the reward for having done it.
The short version
Be readable. A claimed profile, a text menu with prices, hours that match everywhere, a named service area, and reviews that mention actual dishes or treatments. That's what an assistant needs to put you on a three-name shortlist, and it's what a human needs too. If the till integrations reach the UK next year, you'll be ready. If they take longer, you've still fixed the things that were quietly costing you bookings.
Frequently asked questions
Can I pay to appear in ChatGPT's restaurant recommendations?+
No. There is no advertising slot to buy in AI recommendations at present. Assistants assemble answers from indexed web pages, business listings and review sites, so visibility comes from having accurate, readable, specific information published where those systems can find it. Square's July 2026 ordering integration is a till-side connection, not a paid placement.
Is a PDF menu really that damaging?+
Yes, if it's your only menu. A PDF is a separate file that often isn't indexed well, and a photographed menu is just pixels to a crawler. That means an assistant asked for a specific dish cannot confirm you sell it. Publish the menu as ordinary text on a web page and keep the PDF purely for printing.
Do I need Square to benefit from any of this?+
No. Square's integration matters for taking the order inside ChatGPT or Claude, but being recommended in the first place depends on your public information, not your till. Any UK venue can improve its chances today with a claimed Google Business Profile, a text menu or service list, consistent hours and recent descriptive reviews.
How do reviews affect whether an AI assistant recommends me?+
Assistants use reviews for both quality and detail. A high rating helps, but review text that names specific dishes, treatments or circumstances helps more, because it gives the assistant something to match against a request. Encourage customers to mention what they ordered or which service they had rather than leaving a bare star rating.
My hours change seasonally. How should I handle that?+
Update them everywhere at the same time: your website, Google Business Profile, any booking platform and social pages. Conflicting hours across sources make assistants less likely to recommend you, because they cannot state confidently that you're open. Also set bank holiday hours in Google in advance rather than fixing them on the day.
Want a website like the ones we write about?
Describe your business in a sentence and SiteGenie builds you a complete site. Free to try, no card required.
Build my website