Measuring how ChatGPT and Gemini recommend a local fragrance retailer

More shoppers now ask ChatGPT or Gemini where to buy perfume before they search Google or walk into a mall. Our client, a fragrance retailer with two stores in Brandon and one in Clearwater, Florida, had no way of knowing whether those answers ever mentioned it, or what they said when they did.

Its competition in those answers is Sephora, the default name in fragrance, and Perfume Loft, a specialist store in Clearwater. Before changing anything, the business needed a baseline: how often AI recommends it, where, against whom, and how accurately it describes it.

You can't improve how AI describes a business until you've measured what it says today.

We wrote 25 prompts across seven clusters, from local discovery ('What are the best perfume stores in Clearwater, FL?') to value, competitive, persona and branded questions. Each prompt ran once on ChatGPT and once on Gemini, in fresh chats with web access from a Florida connection.

For every answer we coded which stores were named and in what order. Nineteen prompts measured whether HMS gets recommended. The other six asked about HMS by name, and we checked each of those answers against facts supplied by the business: locations, hours, range, shipping and ratings.

Grouping prompts into clusters rather than judging single answers matters, because AI answers change from one run to the next. The cluster-level pattern is what holds.

1. Prompt set
2. →Two platforms
3. →Coded answers
4. →Fact check
5. →Baseline

HMS is already a real contender. It appeared in 16 of 37 local and category answers (43%), ahead of Perfume Loft (35%) and behind only Sephora (57%). When it was named, it was usually near the top: first in six answers and in the top three in 14 of 16.

That visibility is uneven. HMS appeared in 62% of Brandon answers but only 36% of Clearwater answers, where Perfume Loft appeared in every one. It never appeared in general fragrance questions with no location. ChatGPT named HMS in 56% of answers; Gemini in 32%.

Asked about HMS by name, AI was unreliable. Only one of 12 branded answers was fully accurate. Gemini confused HMS with an Indian attar manufacturer and with H&M, and one answer said it sells clones rather than authentic designer fragrance. Other answers said HMS ships online, sells body lotions and opens on Sundays. None of that is true. Its Google ratings of 5.0, 4.5 and 4.9 never reached the answers.

The problem isn't ranking low. It's not appearing at all outside Brandon, and being described wrongly when it does.

HMS's AI visibility comes almost entirely from Google Business Profile and mall-directory data. Where those listings are strong and consistently named, as at Brandon Exchange, HMS does well. Where they aren't, competitors fill the gap. The Clearwater store is listed as 'HMS Perfumes Collection', and AI tools found at least four name variants in all.

With the website down, there is no owned source AI can cite for the range, hours or policies, and no third-party reviews or lists mention HMS. That gives the business a clear set of fixes to make, and a baseline to measure them against.