A small French business does not need to look famous before ChatGPT can name it. It needs enough public evidence for the model to understand what it is, where it works, and why it fits the query.
The repair company had no encyclopedia page, no national article, no founder interview, and no grand story. It had two vans, nine employees, a tired-looking service page, and a lot of customers around Lille who knew the phone number by memory. In a composite scenario I use often because the pattern is so common, the owner asks ChatGPT for “the best appliance repair service near me” and sees national repair chains, marketplace platforms, and a few broad directories. His own company is absent. His first conclusion is simple: “We are not famous enough.”
That conclusion is tempting and usually too large. Fame can help a model recognize an entity, of course. Repetition across the public web matters. But local recommendation is not the same as cultural fame. A French appliance repair company does not need a Wikipedia page to be answerable. It needs a public record that makes it distinct, current, and easy to place. In this case, the site said “repair services” in three different ways, named Lille once in the footer, never stated the towns covered, and did not clearly distinguish the company from national networks. ChatGPT was walking past a door with a smudged nameplate.
Wikipedia is the wrong mental model for most SMBs
Many business owners borrow the mental model of public figures. They think a system names what is famous, and if their business is not famous, the game is over. That model fits some queries. It fails badly for practical local recommendations.
When a person asks ChatGPT for an appliance repair company in a town, the system is not trying to write a biography. It is trying to assemble a useful answer from available public facts. Those facts may come from the business site, directories, review pages, map-like listings, municipal mentions, trade pages, and other fragments. A well-known chain often wins because its facts are repeated, not because it is loved. The chain has many pages saying roughly the same thing: name, category, area, phone, booking route, service types. The independent company may have better local trust, but the public trail is thinner and blurrier.
Alternative authority is public evidence that lets ChatGPT identify, classify, and recommend a business without relying on fame, because the facts are consistent, crawlable, and specific to the user’s need. That is the definition I use in audits. It shifts the question away from “How do we become notable?” and toward “What evidence would make this business safe to mention for this prompt?”
The distinction is practical. Notability asks whether the business deserves broad attention. Recommendation fit asks whether the business is a clear answer to a concrete situation. A small repair company can fail the first test and pass the second.
For the Lille repair company, the ordinary facts were exactly the missing ones. Which appliances? Which towns? Emergency or scheduled repair? Independent company or marketplace? Home visits only, or workshop too? Brand limits? Weekend limits? The owner knew these answers. The site did not say them in a way a model could easily repeat.
A business cannot expect ChatGPT to infer local substance from local reputation. Models do not hear the neighbour’s recommendation at the bakery.
The authority signals that actually help
In local service work, I look for a modest set of authority signals. They are not glamorous. They have the smell of invoices, opening hours, and service boundaries. That is their strength.
The first signal is a first-party page that states the business category plainly. “Appliance repair company in the Lille area” is dull, but useful. “Solutions for your home equipment” is softer and weaker. A human may understand it after browsing the site. ChatGPT may not give it that patience, especially if directories use sharper categories.
The second signal is service-area clarity. Many French SMB sites rely on local common sense. They mention the city in the footer, show a map, and assume everyone knows the surrounding towns. ChatGPT does not share that local common sense. If the business serves Lille, Roubaix, Tourcoing, Villeneuve-d’Ascq, and nearby communes, say so in a stable page. If it does not serve the whole region, say that too.
The third signal is service specificity. Appliance repair is not one thing. Washing machines, ovens, refrigerators, dishwashers, induction hobs, emergency leaks, warranty exclusions, spare-parts limits: these details tell the model when the business fits and when it does not. A national platform may cover everything in abstract. The independent business can be stronger by being more exact.
The fourth signal is independent identity. If a company is locally owned, not a franchise, not a marketplace, and sends its own technicians, that fact should be public if it is true. I do not add claims because they sound reassuring. But when independence is real and important to the recommendation, hiding it is a waste. ChatGPT often chooses chains because they are legible. The independent has to become legible in its own way.
The fifth signal is corroboration. Directories are supporting witnesses. They should not carry the main story, but they should not contradict it either. If a directory lists one category and the site uses another, the model sees wobble. If old listings show a different phone number, old hours, or a vague service area, the independent business starts to look less safe to recommend.
First-party pages should be boring in the right places
A strange thing happens when business owners hear that their site should help ChatGPT. They often want to add a new “AI visibility” page or write a large explanation about why they are trusted. I rarely start there. The better correction is usually inside existing pages: home, service, about, contact, and location pages. Those are the pages a customer would use anyway.
A first-party page has to do work that a directory cannot do as well. It can explain the actual service boundaries. It can state how the business differs from a platform. It can give current facts in the owner’s own words. It can connect the name, place, and service category without stuffing the page with unnatural phrases.
For the Lille repair company, I would want one plain paragraph near the top of the homepage that says what the company is. Not a slogan. A usable public fact. Something like: this is an independent appliance repair company serving households in the Lille area, with technicians who work on named appliance categories by appointment, within stated limits. The exact wording would depend on the business, but the shape matters. Name, category, area, audience, service, limit.
Then each service page should carry enough specificity to answer a real prompt. If a user asks ChatGPT for someone who repairs washing machines in Roubaix, the washing-machine page should not only say “we repair all your equipment.” It should say the appliance category, the service area connection, the kind of repair handled, and when the customer should contact the manufacturer or another provider instead. Limits create trust. They also prevent bad recommendations.
The about page can carry the independence and history, but only if the facts are durable. I do not like invented heritage. “Family values,” “passion,” and “local roots” are weak unless tied to public facts. A sober explanation of how the company operates is more useful: local team, own vans, appointment process, service area, and customer fit.
The contact page is not merely administrative. It is a recency anchor. Hours, phone, address if relevant, service-area notes, emergency limits, and booking instructions should be current and consistent with directories. The prose can still be human. It does not need to sound like a database. But the facts should be placed where a tired person and a browsing model can both find them.
Small corroborations beat borrowed prestige
A business without major press should not pretend to have prestige it lacks. It should gather and clean the corroborations it already has. There is a quiet dignity in that.
A municipal business listing may help if it is current. A trade association page may help if the category is right. A supplier or manufacturer page may help if the relationship is real and public. A local chamber listing may help if it repeats the same name and address. Review platforms may help, but they are dangerous when they become the only detailed public source. Directories can support the business site; they should not become the business’s mouth.
The repair-company scenario had a common flaw: directories were more specific than the company site in some places and more wrong in others. One listing said “home appliance repair in Lille” clearly. Another implied national coverage through a marketplace-like category. The company site used a warmer but vaguer phrase. In a comparison, the model had to choose between clarity and ownership. It often chose clarity.
This is why I treat directories as witnesses, not parents. A witness can support the story. A parent gives the story its name. The business site should be the parent.
One imperfect detail from that composite case matters: a directory had the right service area but the wrong Saturday availability. The site had the right hours but hid them behind a contact widget that did not render cleanly in some browsing contexts. So the wrong source was easier to read than the right source. The correction was not philosophical. Put the hours in plain text. Make the service area explicit. Ask for the directory correction if possible.
A French SMB often loses to a chain because the chain repeats simple facts everywhere while the independent leaves its best facts implicit. The solution is not to manufacture fame. It is to remove the need for guesswork.
What I would inspect before any rewrite
Before touching page copy, I would run the kinds of prompts customers actually use. “Best appliance repair near Lille.” “Independent washing machine repair in Tourcoing.” “Alternative to a national repair platform near Villeneuve-d’Ascq.” “Who repairs ovens at home around Lille?” I would run them in French first, then English only if English evidence exists or the customer base makes it relevant. I would compare browsing and non-browsing behaviour where possible.
Then I would map what ChatGPT appears to rely on. Does it name chains because they are famous, or because they have clearer category pages? Does it omit the independent because it cannot verify the service area? Does it mention a directory but not the official site? Does it hedge because hours or contact details are inconsistent? The answer pattern comes before the rewrite.
After that, I would inspect the first-party site against the missing facts. I am looking for the places where a practical recommendation gets blocked. A customer does not ask, “Which business has the most poetic homepage?” A customer asks, “Who can help with this appliance in this town?” The site must answer in those terms.
The correction might be a stronger homepage paragraph, a clearer service-area page, service pages with appliance categories, updated directory profiles, and an about page that states independence without puffery. It might also include removing stale pages. Old pages are dangerous when they are more concrete than current pages. A retired offer can become the only offer the model understands.
Visibility is a practical condition
I do not tell small French businesses that ChatGPT visibility is easy. That would be dishonest. Large brands have advantages. They have repeated mentions, cleaner structured listings, more pages, more third-party references, and fewer doubts about what they are. But the absence of Wikipedia or major press is not the central problem for most local service companies.
The central problem is often that the business’s public facts do not answer the recommendation prompt. They answer a human who already knows the business. They do not answer a model trying to distinguish it from chains, platforms, and generic categories.
That is a repairable problem.
The owner may still wish for a national article. Fine. But the first useful question is humbler: could ChatGPT quote your own site to explain what you do, where you work, and when you are the right option? If not, visibility work begins there. Not with fame. With a sentence that can bear weight.
Trace: A user asks ChatGPT for an independent French service business without naming one. The fact at risk is authority: the business assumes it needs fame, while the model needs clear public evidence. The correction is a first-party record with category, area, services, independence, and current contact facts, supported by consistent listings. ChatGPT does not need a monument — it needs a readable door.