AI
Edward Ilin12 min read
A shopper asks a chatbot for a road bike. The model hands back a list of national chains. The independent shop around the corner never appears.
Your small-retail clients are being edited out of the answer.
Search Engine Land, reporting on a study for Lightspeed Commerce, found large retailers were the top recommendation in 46% to 58% of AI shopping responses even with live search on.
Here is the problem you are managing this quarter. Your clients' content calendars were built for search results pages, not chatbot answers. You are already stretched across too many accounts, and spending hours on volume that AI systems ignore is the worst use of your week.
You can still move your clients into that answer without outspending a chain, and the fix is narrower, cheaper and faster than another round of blog posts.
Discovery used to be an SEO question: do you show up in Google results? Search engines learned a shopper's location and surfaced nearby stores. Chatbots answer differently. They decide what to recommend, and that decision leans toward the biggest names.
The evidence is consistent. Search Engine Land reported on a Vaer AI study for Lightspeed Commerce and said that even with live search, large retailers were the top recommendation in 46% to 58% of responses across ChatGPT, Google AI Mode and Google AI Overviews. An Entrepreneur contributor piece drawing on that research says that nine times out of 10, AI directs shoppers to major retailers.
The two figures measure different things. Search Engine Land's range covers live-search conditions; the nine-in-10 claim is the Entrepreneur contributor's own account. Repeat the range to clients.
The cost side is just as pointed. A Google-surfaced item on Walmart's $1 trillion AI transformation notes that for smaller regional retailers, the cost of entry for this level of AI integration is becoming a barrier that may lead to further market consolidation.
A retail restructuring roundup surfaced through Google says analysts at UBS project that as many as 40,000 U.S. stores could close over the next five years. That is a projection, not a verdict on any one client. Still, it is the backdrop for every retainer you hold with a small retailer.
A client whose organic traffic erodes does not blame the model. They question the retainer. Product-name searches are the sharpest edge: the Entrepreneur piece says that when a shopper names a specific product, a small store's odds of being the top recommendation fall from roughly one in three to one in 10.
Do not try to match a national chain on catalog scale. You will lose, and your client will pay for the attempt. Posting more generic buying guides feeds the exact content AI can already assemble from larger, better-structured sources.
What a chain cannot fake is a specific store's knowledge. Which tires suit the local hills. Which fit problems come back every spring. What a regional customer actually reported after buying. That material is slow to produce, which is exactly why it holds value.
Pitch scarcity, not volume.
The Entrepreneur contributor's advice is to establish authority around what a store sells. A bike shop, for example, can publish short videos on setting up tubeless tires or replacing a worn chain. Those clips look educational, but they demonstrate the expertise behind the business.
Check YouTube first. The same piece notes that Gemini regularly surfaces videos from the site, and that small luxury watch dealers use long-form videos to build trust on topics like avoiding fakes. Social posts now double as a signal to AI about what a business sells and specializes in.
Reformatting eats your hours. A tool that learns a brand's voice once and turns one expert idea into a week of posts, articles and video scripts, such as Brainpercent, lets you spend your limited hours collecting the proof instead.
Most client briefs still ask for a keyword, a word count and a ranking goal. Rewrite them to name the question a shopper would put to a chatbot, then specify the proof that should answer it.
Add structure for machines. The Entrepreneur piece recommends clearly labeling product details such as price, brand and stock status, and giving search and AI crawlers a map of the pages you want found. AI agents increasingly fetch product information directly, so a retail site works as a showroom for both software and people.
One more client lesson from that piece: coach shoppers too. Adding words like "independent" or "locally owned" to AI searches cut big chains' share of recommendations to as little as one in 10 on Google's platforms, according to the study the contributor commissioned. Clients can share that tip in-store, on their sites and across social.
Pick the three accounts where traffic depends most on product-name searches. Open their top product pages and read them as an AI model would.
The output is a short list per client: absorbable pages and defensible pages.
Do one, not three. Choose the client with the clearest expertise and build a single page around a question shoppers actually ask.
Put first-hand material on it: a real fitting scenario, a regional inventory note, a customer outcome in the owner's voice. Pair it with a short video if the client can film one. Treat it as a test case you can show the other two.
Do not wait for a traffic drop to start the conversation. Send each exposed client a short note with the Search Engine Land finding, the product-name risk, and your recommendation.
A client who hears it from you now is a client who renews.
Start with a test the client can watch happen. Open ChatGPT or Google AI Mode on a shared screen, type a shopper question for their category with their city in it, and read the answer aloud. Then run the same question again with "independent" or "locally owned" added. If the client's shop shows up in the second run and not the first, you have just shown them the gap better than any slide deck could. Screenshot both answers and keep them for the quarterly review.
Then separate two things the client may be blending together. Their existing posts may be perfectly good. The issue is that a model deciding what to recommend favors sellers whose catalogs, stock and reviews are easy to read at scale.
Frame your plan as adding the machine-readable layer and the first-hand proof on top of what they already publish, not replacing it. That keeps the conversation away from blame and toward a short list of fixes.
Build a small, repeatable prompt set per client. Ten to fifteen questions is plenty: five broad category questions, five that name a specific product, and a few with a local modifier. Run them on a fixed day each month across the same assistants, and log three things in a simple sheet: whether the client appears, whether a national chain appears instead, and which sources the answer cites. Keep the wording identical every time so you are comparing like with like.
Expect noise. Answers change from one run to the next, so one miss means little and a trend across a few months means more. Treat the sheet as a directional signal for the client conversation, not as a guaranteed metric.
The citations column is the most useful part. It shows you which pages and videos the models actually pull from in that category, and that tells you where to put your next piece of proof.
Rank accounts by two things: how much of their revenue comes from branded product searches, and how much original expertise they can give you. A shop that sells the same manufacturer lines as every chain and has an owner who never films anything is your most exposed account with the least raw material. A shop that does repairs, custom fits or sourcing has material you can work with right away. Put your first hours into the exposed accounts that also have something to show.
For the exposed accounts with thin material, the first job is simpler: get their inventory and product details cleanly labeled and online. That work does not depend on the owner's time, so you can finish it while you wait on footage.
Lower the cost of capture. Skip the polished shoot and ask for a fifteen-minute phone call or a voice memo recorded between customers. Keep one question per call, such as "what do people get wrong when they buy this?" or "what comes back to the repair bench every spring?" Record it, pull out the specifics, and have the owner approve the final wording. One call can feed a page, a short video script and several social posts.
Keep a running file of customer outcomes too. A single sentence from a buyer, shared with permission, carries more weight than a paragraph of general advice. Send the owner a recurring two-line prompt each week asking for one story from the shop floor. Owners who will not sit for an interview will often answer a text, and a text is enough to start from.
Pair every warning with a priced, small first step. The reported concern is that the cost of AI integration is becoming a barrier for smaller regional players. According to a Google-surfaced item on Walmart's AI transformation, that barrier may push further market consolidation. Make clear that your plan does not require matching a chain's spend. It asks them to fund a handful of proof pages and a clean inventory feed, which a chain cannot copy store by store.
Give the client a ceiling and a checkpoint. For example, propose a fixed number of proof pages over a set period, then review the prompt-set results together before committing more. A client who sees a bounded cost and a scheduled review is far less likely to freeze or walk. Never promise a specific lift in recommendations, because the models change and nobody can guarantee one.
Start with the audit today: open one top product page and read it the way an AI model would. You cannot match a chain on catalog size; what it can't fake is a store's own knowledge: local fit advice, real stock, customer results in the owner's words. This week, audit three clients, build one proof page, and send each a short note. Pitch scarcity, not volume.
Ready to automate all this? Brainpercent is the all-in-one content platform that generates SEO articles, social posts, and videos for you, on autopilot. Start your free trial or see pricing.
Posts, images and videos from your own site, in minutes.
Try it free