Category, brand, place, condition: a monthly reading for one store

What a store can be named for

Shoppers ask for a store by what it sells. "Where can I buy a kids' bike in Calgary" asks for a category in a place, and "who carries [brand] trail shoes near me" asks for a brand. Add "open Sunday", "in stock in a size 10" or "where I can try them on" and the question also carries a condition, something that has to be true on the day the shopper comes in.

An answer that names your store has named it for one of those parts, and the part matters. Named for the category, you are one of the places that sell kids' bikes. Named for the brand, you are where that maker's line can be found. Named for a condition, the answer has said something about your hours or your stock that a shopper may act on, and that something can be wrong. A reading of a store's AI visibility keeps the parts apart, because each one points at a different page.

Named, recommended, first: three readings of one answer

Three things are recorded for every answer. Named means your store's name appears in it anywhere: in a list, in a sentence, beside a link. Recommended means the answer puts your store forward for the shopper's need, with or without a reason ("try [store name], which fits shoes by appointment"). First means yours is the first store the answer names, in a list or in running prose.

The three come apart. A store can be named without being recommended, as one name in a list of five, and recommended without coming first. It can also be named for something that is no longer true: "[store name] carries [brand]" in a month when you have stopped ordering that brand. Record that answer as named, and mark the fact wrong. It sends a shopper to your door for something that is not on the shelf, which is why the accuracy columns below count for as much as the name.

Our methodology builds a visibility score from 0 to 100 out of the first of the three alone, the share of the twelve questions that named the store, with no weighting. The score tracks one store against itself from one reading to the next. The columns tell you which page to change, and an AI recommendation that carries a wrong brand or a wrong closing time is a finding before it is a point on a chart.

A panel by category, brand, place and condition

The panel is a fixed list of prompts, written the way your shoppers type and put to ChatGPT, Claude, Gemini and Perplexity every month. For a store, build it from the four parts: a few category questions for your place, the brands you most want to be found for, the conditions your shoppers attach, and two questions about the store by name. The template below has twelve lines. Fill in the brackets once, then leave the wording alone.

Write each line the way it would be typed, lowercase and loose, and keep it that way. How to hold the panel as a file, why every question runs three times on each assistant, and when a line may be replaced are all in the article on prompt monitoring; the method is the same for a store as for any other trade. Only the brackets are yours. A store that rents nothing drops the rental line and adds a second brand.

What to write down beside each answer

The standard record says whether you were named, recommended and first, and who else was named. A store adds four columns: one for what it was named for, and three that hold the answer against the store as it stood that day. Keep the brands page, the hours and the stock list open while you read, because right means right on the reading day, not right last season.

Two habits make the columns usable later. Keep each answer verbatim beside its row, so a wrong fact can be traced back to where it is written. And list every other store the answer names, in order: that list is the raw material for share of voice, and it shows when a competitor is named for a condition, such as open Sunday, that you meet as well.

Reading the first month, pattern by pattern

The first month is a baseline, not a verdict. Read it by pattern rather than by score, since each pattern points at a page:

• Named for categories, never for brands: check that the brands page exists, is dated and matches the shelf. The same-day change routine sets out a dated brands page.
• Named with hours that were wrong on the day: compare the site, the Business Profile and the markup for that day; the same routine keeps them in step, holiday and seasonal hours included.
• Named for the place, never for stock or pickup: check whether any page states those facts as text. Availability and pickup, stated on the page covers them.
• Never named for the fittings, setups or repairs you do: check whether the advice is written down anywhere, which a specialist store's advice, in writing deals with.
• A fact about you cited to a directory or another store's page: check whether your own page states it. The afternoon check against the stock list shows how to trace a wrong fact to its source.

Pick one or two patterns to work on and leave the rest for a later month, so that a change in the answers can be set beside the change you made. The guide for retail stores lists the pages in the order the work goes.

What the reading cannot tell a store

It does not count visits or sales. A shopper who reads your name, drives over and pays at the register leaves no trace in the panel. If you want that count, ask new customers at the counter how they heard of you, from a short fixed list that includes an AI assistant, and keep the question the same every month.

It does not say why an answer came out as it did. The operators do not publish how their assistants compose an answer, and the record holds what was said, not the reason for it. Answers also vary from run to run, so one question that flips between two months may be noise. What the reading is for is direction over several months, on questions that have not changed.

And it sees only its own twelve questions. When the store takes on a new category or a new brand, add a line for it with its own start date. Never reword an existing line to make room, because a reworded question starts a new series.