Local & Industries · 9 min read · July 15, 2026
Am I being recommended? Making AI visibility measurable in retail
More and more people no longer ask Google but ChatGPT, Gemini or Perplexity: "Where do I buy the best running shoes in my city?" Whether your shop appears in that answer decides real revenue. This guide shows you how to measure, understand and deliberately improve your AI visibility in retail, step by step.
Why the question "Am I being recommended?" decides your revenue today
Imagine someone standing in your city in the evening, typing into ChatGPT: "Where can I still get a gift for my mother here today that doesn't look like standard fare?" The AI answers with three, maybe four specific shops. Are you among them or not? That is the new starting question in retail. The customer no longer compares ten blue links but reads a finished recommendation and then often sets off directly. Whoever gets named wins the visit almost in passing.
For you as a retailer this means a real shift. In the past you fought for rank one on Google; today you fight for a place in a list the AI compiles itself. This list is shorter, harder and less transparent. Whoever isn't named simply doesn't exist for this customer. And unlike on Google, no statistic shows you that you were just passed over. There is no missing click you could count.
That is the core of the problem: AI recommendations are invisible unless you actively measure them. A fashion retailer in Regensburg can believe for months that everything is going well, while ChatGPT consistently names three competitors for "nice boutiques in Regensburg" and never them. That is why the first step isn't optimisation but honest measurement. Only once you know where you stand can you change anything at all.
What AI visibility concretely means in retail
AI visibility means: does your shop appear in the answers of generative AI systems when someone asks about your range, your location or your category? This concerns ChatGPT, Google Gemini, Microsoft Copilot, Perplexity and, increasingly, the AI overviews directly in Google Search. These systems draw their answers from web content, directories, reviews and structured data. They form a judgement about you from that, whether you want them to or not.
The difference from classic search engine optimisation is important. With SEO the point was to rank visibly. With GEO, Generative Engine Optimization, the point is to be correctly understood by the AI and actively recommended. A bike retailer doesn't just want to be found, they want the AI to say: "For e-bikes with a good workshop, shop X near you is a good address." That's a recommendation, not a mere listing.
For brick-and-mortar retail there is an added twist: the local reference. Most relevant queries contain a place or a "near me". Your AI visibility therefore depends strongly on how clean your local data is, what your reviews look like and how unambiguously the AI can assign your range to a city or district.
The first step: define your own prompts
Measurement begins with the right questions. Sit down and write out twenty to thirty prompts that your real customers would ask. Don't think like a marketer but like a person with a need. Examples for a sports specialist shop: "Where do I buy running shoes in Freiburg with professional advice?", "Which sports shop in Freiburg has a good selection for hiking?", "Where can I get club jerseys printed?"
Mix three kinds of prompt. First, category prompts without a brand, such as "best wine shop in Mainz". Second, occasion prompts that describe a problem, such as "gift for a coffee lover under 30 euros in Mainz". Third, brand prompts with your name, to check whether the AI even knows you and describes you correctly. The last type in particular often uncovers embarrassing errors, such as wrong opening hours or a long-since-changed range.
This prompt list is your measuring instrument. It should reflect your actual business, not your wishful thinking. A deli that really lives off its cheese counter but only tests prompts for "deli in general" is measuring past reality. The more concrete your prompts, the more honest the result.
Measure systematically instead of trying it once
Many retailers type their name into ChatGPT once, are satisfied or shocked, and then leave it. That isn't measurement, that's a snapshot. AI answers fluctuate: the same question can name your shop today and not tomorrow. So you need a system. Run each of your prompts across several AI systems and on several days, and each time note whether you're named, in what position and how you're described.
Enter the results in a simple table. Columns: prompt, AI system, date, named yes or no, position, tone of the description, competitors named. After two or three runs a pattern emerges. Suddenly you see in black and white that you never appear for category prompts but are described correctly for your name. That's a clear diagnosis: the AI knows you but doesn't assign you to your category.
Over the weeks this table becomes a curve. This visibility rate, the share of prompts for which you get recommended, is your most important metric. It replaces gut feeling with a number you can improve. Whoever measures can prove progress. Whoever only guesses only argues.
The most common blind spots in retail
When you evaluate your results, you usually hit the same causes. The first classic is contradictory data online. Your Google Business Profile says "open until 8 p.m.", your website says "until 6 p.m.", an old trade directory names a branch you gave up long ago. The AI stumbles over these contradictions and, in case of doubt, becomes cautious, so it prefers not to name you at all rather than claim something wrong.
The second blind spot is a range that is nowhere in text form. You've stocked sustainable outdoor brands for a year, but your website only says "clothing and accessories". The AI can only recommend what it can read. If your unique selling point only hangs in the shop window and isn't on an indexable page, it doesn't exist for the machine. It is exactly here that many brick-and-mortar retailers lose their chance.
The third point is reviews. AI systems read reviews too and derive descriptions from them. Few, old or one-sided reviews give the AI little material to classify you positively. A shop with thirty fresh, content-rich reviews praising advice and selection gives the AI exactly the phrasings it will later carry forward.
From measurement to improvement: concrete levers
Once you know your blind spots, optimisation gets very concrete. First bring your base data in order: identical address, opening hours and phone number across website, Google profile and all directories. This consistency sounds trivial but is the foundation for the AI to trust you at all. A single clean data record often works more strongly than any ad campaign.
Then write your range and your strengths in clear, searchable text. Create pages that answer concrete questions: which brands you stock, for which occasions you're the right shop, what distinguishes your advice. Phrase it in full sentences, the way a customer would ask. A page "Running shoe advice in Freiburg" with real content moves you forward for exactly such prompts, because the AI can read the connection directly.
The third lever is structured activity: regularly gathering reviews, keeping your offering current and being mentioned in local contexts, such as city magazines or trade listings. Every clean, consistent mention on the web is another piece of evidence the AI can draw on. After a few weeks you repeat your measurement and see on the curve whether the levers are taking hold.
Competitive comparison: who gets recommended instead of you?
Your measurement table contains an often-overlooked gold seam: the names of the competitors the AI names instead of you. Look at these shops closely. What do they do that you don't? Usually you'll find a well-maintained website with clear category texts, many recent reviews and unambiguous local positioning. That's no coincidence but the result of exactly the signals the AI rewards.
Use this comparison not for frustration but as a template. If the same competitor always appears for "sustainable fashion in Cologne", analyse how they describe their offering and where they're mentioned. You don't have to copy them, but you understand which gap the AI still sees in you. Often it's enough to catch up on one or two of these signals in a targeted way to slip into the recommendation yourself.
Keep an eye on the comparison over time. AI visibility isn't a one-off state but a moving field. Competitors change their presence, AI systems update their data basis. Whoever measures and compares regularly notices shifts early and can react before customers migrate away permanently.
Your next steps this week
You don't have to tackle everything at once. Start small and concrete. Take on three tasks this week: first, write out your twenty most important customer questions as prompts. Second, test them once through ChatGPT and Gemini and enter the results in a table. Third, check whether address, opening hours and range match on your website and in your Google profile. These three steps alone give you more clarity than months of guessing.
In the second week you tackle the content. Write real text pages for your two or three most important categories that answer concrete customer questions. Actively ask satisfied customers for a review and give them a reason to mention your advice or selection. This work pays into your visibility slowly but steadily.
And then keep measuring. Set a fixed date every four weeks on which you go through your prompt list again and note the visibility rate. This turns a diffuse feeling into a controllable metric. The honest answer to "Am I being recommended?" may be sobering at first. But it's the beginning of taking it into your own hands.
Common questions
Isn't it enough just to rank well on Google?
No, today that's only half the battle. More and more customers get their purchase recommendation directly from ChatGPT, Gemini or Perplexity, without ever landing on a classic search results page. These systems partly draw from the same sources as Google, but weight them differently and output only a few names. You can be on page one of Google and still be missing from the AI recommendation. That's why you should measure both separately.
How often should I measure my AI visibility as a retailer?
A good rhythm is a thorough measurement every four weeks. AI answers fluctuate from day to day, so a one-off test is worthless. Run your fixed prompt list across at least two AI systems and note the results. If you're currently working actively on your website or your reviews, a tighter cadence pays off, to see whether your measures are taking effect. What matters is regularity, not frequency in itself.
My shop doesn't even have a big website. Do I have any chance at all?
Yes, but you have to give the AI material. Even without an elaborate website you can move a lot: a cleanly maintained Google Business Profile with correct data, recent and content-rich customer reviews, and entries in relevant local directories. Add a few clear text pages about your range and your strengths over time. Small, specialised shops in particular often have a strong profile the AI likes to recommend, as soon as it's available in readable form on the web at all.
Read on