Google Maps Will Soon Place Food Orders on Its Own. Restaurants Need Menus an AI Can Actually Read.

On August 6, 2026, Google announced that Ask Maps — the conversational assistant built into Google Maps — can now do more than answer questions about a restaurant. A customer can type or say something like “order spicy pad kee mao with seafood for me to pick up on my way home,” and Ask Maps will find an open restaurant along their route, check it against any dietary needs mentioned, and build a cart on the restaurant's own ordering system before handing it back to the customer to pay. The feature is rolling out in the US first, through point-of-sale partners Square and Toast, with Uber Eats support coming soon. Google also said it is co-developing a Universal Commerce Protocol for Food with its partners — an open specification for how an AI agent is supposed to read a restaurant's menu and assemble an order, rather than just display it to a person.

None of this is live in the UAE yet, and it may be a while before it is. But the detail worth sitting with isn't the checkout button. It's the requirement sitting underneath it: an AI agent placing an order on a customer's behalf has to actually parse the menu first — the dish names, the prices, what's in stock, what's spicy, what has seafood in it. It cannot do that from a photograph of a laminated menu card or a PDF scan sitting behind a QR code, which is exactly what a large share of restaurant “digital menus” still are today.

Why this matters even before agentic ordering reaches the UAE

You don't need Ask Maps to be live here for this to be a useful prompt to check your own setup. A menu that's really just a photographed PDF behind a QR code already has real costs with human customers: it's slow to load on a restaurant's WiFi, the text is often too small to read without pinching and zooming, prices can't be searched, and updating a single item means reprinting and re-uploading the whole file. Those are today's problems, and they exist regardless of what Google's assistant can or can't do yet.

What the Ask Maps announcement adds is a second, forward-looking reason to fix the same thing. The direction is unambiguous: ordering technology, starting with Google and almost certainly followed by others, is moving toward machines reading structured menu data directly, not screenshots of paper. A restaurant whose menu already lives as clean, structured, dashboard-managed data — items, categories, and prices as actual fields rather than pixels in a scanned image — is simply better positioned for wherever this goes next, in whichever market it reaches first. A restaurant whose only “digital menu” is a photo of the same laminated card from three years ago is not.

What to actually check this week

You don't need to chase Google's roadmap to act on this. Two honest questions are enough:

• Is your menu data, or is it a picture of your menu? If updating a price means reprinting a QR sign or re-scanning a PDF, it's a picture. If you can log into a dashboard and change a price in thirty seconds, it's data.
• Can a customer search or filter it on their phone in under three seconds? If they have to zoom and scroll through a static image, most won't bother — and neither would an automated agent, if one were reading it.

This is exactly the gap Smaacaad's Digital Menu Card (AED 29) is built to close, and it's worth being precise about what it does and doesn't claim: it does not connect to Google's Universal Commerce Protocol, which doesn't exist as a public standard yet. What it does is put your logo on one side of a card and a QR code on the other that opens a real digital menu — items, categories, and prices entered as structured data on a dashboard at dm-only.com, not a photographed file. Guests tap the card with an NFC phone or scan the code, no app required, and when a price or dish changes, you update the dashboard instead of reprinting anything. That's the same underlying discipline agentic ordering will eventually depend on, and it happens to be the same thing that makes the menu faster and easier for a human to use today — which is the actual reason to fix it now, whether or not an AI agent ever reads it.

Photo by Rodrigo Rodrigues | WOLF Λ R T on Unsplash