METHODOLOGY: AI DIAGNOSIS

How the AI Visibility Check is measured

NIXUMA's AI Visibility Check asks several AIs repeatedly, using questions written the way guests actually ask, and measures three things: (1) whether your restaurant is named in the answer, (2) which restaurants are named instead, and (3) whether anything said about your restaurant does not match the facts. Results are reported as a share (how many times out of how many your restaurant appeared), not from a single answer. Because AI answers vary from one ask to the next, our principle is never to treat a one-off result as a “score.”

WHAT WE MEASURE

What we measure: four things

01 Appearances (named in the answer)We ask the way a guest asks an AI to find a restaurant: “area + cuisine,” “area + occasion (anniversary, business dining, family meal, and so on),” and count whether the restaurant in question is named in the answer. We use both English questions, written as visitors would ask, and Japanese questions, written as local guests would ask.

02 Which restaurants appeared insteadWhen your restaurant does not appear, we record which restaurants the AI recommended instead. If you do not know who you are losing to, you cannot know what to fix.

03 Descriptions that do not match the factsWe check whether details in the AI answer (opening hours, closing days, location, price range, cuisine) match your official information. Wrong details cost visits directly.

04 Patterns in the sources the answers draw onWe look at which kinds of public information the AI answers appear to rest on: your official site, your Google listing, review sites, and so on. When the sources lean one way, that lean itself points to what to fix.

PRINCIPLES

How we measure: five principles

01 We measure across several AIsYour guests do not all use the same AI. A result from one AI alone is not “how you appear in AI.”

02 We repeat, and report sharesAI answers change from one ask to the next. We run the same questions multiple times and report with the denominator: “appeared 3 times out of 10.” We never publish a standalone score calculated from a single result.

03 We ask in your guests' wordsOur questions reproduce how real guests ask. We never measure with questions that include the restaurant's name (“Is X any good?”). The standard is how a guest who does not know your name would ask.

04 We keep the raw answersWe record the exact questions we used and the AIs' answers themselves, so you can check them alongside the report. A summary alone cannot be verified.

05 We guarantee nothingThis measurement measures how you appear now; it does not promise future appearances. We publish the method so you can judge us on the honesty of the measurement, not on whether the result is flattering.

LIMITS

The limits of this measurement

AI answers can change from day to day, even for the same question. And answers obtained through a program may not match exactly what a guest sees in an app. That is why our premise is not to treat a single point in time as absolute, and instead to repeat the same method and watch the change. The AI Visibility Check is a light measurement for understanding where you stand; the Full Audit raises the number of questions, the range of AIs, and the repetitions to increase precision.

This method is not yet final

We will be honest about this. The final measurement items will be settled based on the results of the first check we are now running on NIXUMA itself. Whatever those results are (good or bad), we will publish them on this site as they are. If items or aggregation change, we will update this page and leave the changes visible.

We will not publish unsettled numbers (such as weighting ratios) first to make the method look “scientific.” What we can publish is what is written here: the principles and the measurement items.

For the full service overview, see the AI Visibility for Restaurants page. Questions are welcome at [email protected].