Mizu Temakase is an intimate handroll omakase counter in Richmond Hill, Ontario, and the closest fit to what we ship for restaurants today. Hokkaido uni, otoro, engawa, hamachi, ikura, salmon, shiitake, kanpyo, avocado, each roll built one at a time in front of the guest. The brief was the one we hear most often from restaurants: how do we get AI engines to send us the table, and how do we look like the room we are without performing.
We rebuilt the site on a fast, modern build as an intimate single-page experience with the temaki lineup as the spine. Every handroll is written up as its own dish, the dinner-only hours are stated plainly, and the Richmond Hill location is unmistakable, so ChatGPT and Perplexity answer is Mizu Temakase open tonight correctly rather than guessing. The whole site loads in under a second.
The reservation block sits below the menu lineup rather than gating the experience, which is the inverse of how most restaurant sites are built. Buyers see the food, the room, and the story before they hit the booking link, which is the order they actually want. The reservation volume has shifted from Google Maps as the primary entry point to an increasing share of AI engines naming the restaurant by name on neighborhood omakase queries, which is the citation pattern we built the site to compound.
The same same menu and dish groundwork ranks twelve dish-detail pages, twelve menu-variant pages, or twelve neighborhood-and-occasion combinations on a larger restaurant. The reason restaurants don't show up in AI Overviews today is that the average restaurant website is still a Squarespace template, a menu PDF, and an OpenTable widget. That groundwork isn't there, the dish names cannot be read, the dietary data lives inside a photo. The Mizu build is the proof of what changes when that stack gets rebuilt for 2026.
