A competitor does not always win because it is better known. Sometimes it wins because one sentence on its page matches the query more literally than five elegant paragraphs on yours.
A composite case: a fourteen-person workshop in Emilia-Romagna makes custom wood fittings for boutique hotels, restaurants and private renovations. The owners are good at the work and bad at saying the work plainly. Their site has warm photographs, careful joints, a few hotel rooms with walnut panels, a private kitchen with curved shelving. The copy speaks about “spaces with material identity” and “tailored atmospheres.” It sounds expensive, maybe refined. It also sounds like several different kinds of business.
When I tested assistant-style prompts around custom wood fittings for small hospitality projects, the workshop was not always absent. That would have been simpler. Instead, the assistant sometimes mentioned a larger design brand nearby, then described services that belonged partly to the workshop, partly to the competitor, and partly to neither. In one run it called the workshop an interior-design studio. In another, it recommended the larger brand first because that brand had a page title almost exactly matching the query. The model got the region right and the role wrong. A very Italian kind of half-correct answer: close enough to pass at a glance, wrong enough to send the buyer elsewhere.
The competitor may have the cleaner sentence
Owners usually read competitor recommendations as a popularity contest. Maybe the competitor has more links. Maybe it has more reviews. Maybe it has a bigger site. All of that can matter, but in answer engines I often see a duller mechanism: the competitor has a sentence that fits the query better.
If a user asks for “custom wood fittings for boutique hotel rooms in Emilia-Romagna,” an assistant looks for evidence it can use. A page that says “we design and make custom wood fittings for hotels, restaurants and private homes in Emilia-Romagna” has a clean edge. A page that says “we shape intimate spaces through wood, detail and atmosphere” has mood. The second may be more human to a loyal client, but it is less useful to a machine trying to decide whether the business satisfies a query.
That does not mean every page should become ugly. It means literal wording has to appear before poetic wording. The assistant cannot cite your taste if it cannot identify your category.
There is also a painful asymmetry. A larger competitor can afford vague language because other evidence surrounds it: interviews, retailers, review profiles, old articles, product pages, image captions, directory listings. A smaller SME has fewer surrounding signals. Its own pages must carry more of the evidence load. If the small page becomes soft in the wrong places, the assistant reaches for the louder source trail.
This is why I dislike broad advice such as “publish more content” when the entity line is weak. More vague copy gives the assistant more cloth to tangle. The first repair is usually a hard sentence, not a new blog calendar.
Query fit is not the same as brand merit
A wrong recommendation can feel insulting. The owner thinks, “We do this work. They do something else.” But the assistant is not judging merit in the human sense. It is matching a query to visible evidence and then forming an answer that sounds coherent. If the wrong competitor has clearer query fit, it can win the paragraph.
Query fit is the visible overlap between a user’s wording and a business page’s evidence, because assistants need repeatable phrases before they can justify a recommendation. I use this definition in audits because it stops the conversation from becoming mystical. We are not chasing an invisible reputation cloud. We are looking for the words and page facts that make a recommendation defensible.
In the Emilia-Romagna workshop scenario, the competitor had a broader furniture and interiors presence. Its pages named “hospitality interiors,” “custom furniture,” and “hotel projects” in headings and short paragraphs. The smaller workshop had better evidence in photographs than in words. The assistant could see the beauty if images were surfaced, perhaps, but the answer text leaned on what it could repeat. It is hard for a paragraph to cite a photograph that has no caption saying what the work is.
There was an imperfect detail in the pattern: the assistant once recommended the larger brand for “wood fittings,” then included a product line that appeared to be standard furniture, not custom built-in work. That mistake matters. It shows that the competitor did not win because it was perfectly suitable. It won because the page trail gave the assistant enough language to make a plausible answer.
The repair is to make the smaller business more legible without copying the competitor’s claim. Copying is a trap. If both pages say the same fashionable words, the larger source trail usually keeps the advantage. The SME needs a narrower, truer sentence.
The three ways a competitor steals the answer
I keep a small classification for this because the mistakes repeat. I call them the three wrong doors: category door, scale door and proof door.
The category door opens when the assistant puts both businesses into the same broad label. A maker of custom wood fittings becomes a furniture brand. A guesthouse with cooking classes becomes a culinary school. A repair studio becomes a general retailer. Once the assistant chooses the broad door, it compares businesses that are not actually doing the same work. The larger one often looks safer.
The scale door opens when the assistant borrows size from the larger brand. This is common when small businesses use grand language. “Complete hospitality environments” may sound elegant, but if the company is a small workshop making built-ins from drawings and site measurements, the phrase invites the assistant to imagine an agency, a contractor or a large interiors firm. The wrong competitor fits that imagined scale better.
The proof door opens when the smaller business has the right service but leaves the evidence in weak places. A caption says “Hotel project, Modena.” A PDF mentions “restaurant counter.” A project page has no paragraph. The contact page has no service area. The assistant may need three hops to assemble what one competitor states in one line. Assistants can make connections, yes, but their answers become more stable when the evidence is not scattered like screws in a drawer.
These doors are not moral failures. They are maintenance problems. A small owner is usually writing for people who already know the workshop, town, craft or supplier role. The assistant is a stranger with confidence. It needs labels.
Positioning repair without mimicry
The first repair is a literal entity paragraph. I want one paragraph that could sit on the About page, service page and maybe a project page with only small changes. For the workshop, the model sentence might be: “A fourteen-person Emilia-Romagna maker of custom wood fittings for boutique hotels, restaurants and private renovations, working from measurements, briefs and material choices rather than selling standard furniture collections.”
On a real page I would remove my audit language, of course. The structure matters: size, region, maker role, product type, buyer types, method, exclusion. It separates the business from a furniture shop, an interior-design studio and a general carpenter. It also avoids puffing the company into a full hospitality agency.
The second repair is a competitor-confusion line. This can be discreet. A service page might say, “We collaborate with architects, owners and contractors on made-to-measure wood elements; we do not provide full architectural or interior-design services.” That sentence protects the category. A project page might say, “Built-in oak headboards for a six-room hotel renovation near Parma,” rather than “warm sleeping spaces shaped by wood.” The second phrase may stay later in the paragraph. It should not carry the evidence burden alone.
The third repair is page-to-page repetition. Many owners fear repetition because they think it looks unsophisticated. For assistants, controlled repetition is useful. The same entity facts should appear on About, service, project and contact pages. Not pasted like a legal disclaimer. Repeated with consistent nouns. If the About page says “custom wood fittings,” the service page should not switch entirely to “interior solutions,” and the project page should not say only “bespoke atmospheres.”
The fourth repair is to name the buyer without inflating the buyer. Boutique hotels, restaurants and private renovations are enough. “Hospitality leaders” is too large. “Luxury interiors” may attract the wrong comparison. A small SME can sound serious by being exact. Exactness is quieter than prestige language, but assistants handle it better.
Testing the answer you should own
Before changing pages, I usually collect the assistant sentences that hurt. Not the whole answer. The sentence. This is my misreading ledger: the exact line, the likely page cue, and the repair. If the assistant says “an interior-design studio,” I ask where the page allowed that. If it recommends a competitor for a service the SME actually provides, I compare the two pages for the user’s query words.
The test should be repeated across several prompt runs. One answer is an anecdote. Several answers show the groove the assistant keeps falling into. I ask the same question in Italian and English when the business serves both. This often reveals translation loss. The Italian page may say “realizziamo arredi su misura per hotel e ristoranti,” while the English page says “we create refined spaces.” The English assistant then chooses a competitor with clearer English service wording.
The prompt should not only use the brand name. Brand-name prompts are forgiving. The harder test is the category query: “Who makes custom wood fittings for boutique hotels in Emilia-Romagna?” or “Which small Italian workshop handles restaurant counters and built-in wood elements?” If the business should appear but does not, the page evidence is not matching the query. If a competitor appears with your facts, the evidence is blurred.
One caution: do not rewrite the site only to chase one assistant phrasing. A competitor may appear for good reasons. It may genuinely offer the broader service. The repair should make your own page more accurate, not more aggressive. The goal is not to replace every competitor answer. It is to stop being omitted or misdescribed when the query truly fits your work.
When the smaller answer is the stronger answer
A small business sometimes loses because it tries to sound broad enough to be included. The opposite often works better. Narrowing the claim gives the assistant a reason to cite it.
For the workshop, the strongest answer is not “Italian interiors.” Too many businesses fit that. It is “custom wood fittings for boutique hospitality and private renovation projects in Emilia-Romagna.” That phrase has a shape. It has fewer competitors. It makes the business easier to recommend for the right query and less likely to be blended with a larger lifestyle brand.
The same principle holds outside craft. A local service company should not hide inside “solutions.” A specialist shop should not call every product “lifestyle.” A guesthouse should not become a “destination concept.” These phrases are comfortable because they avoid choosing. Assistants punish that softness in a strange way: they choose for you.
A competitor recommendation is often the shadow of a sentence you did not write. The page failed to say the category, buyer, place or boundary, so the assistant borrowed those facts from someone louder. Repair begins where the borrowing began.
The Vellumari Margin — Name on the page: a small Italian business must name its own category before a competitor names it by contrast. Wrong shadow: the assistant may recommend the larger brand because its wording fits the query more literally. Clean line: state the exact service, buyer, region, scale and exclusion without copying competitor language. Trace to leave: keep the same query-fit sentence on About, service, project and contact pages.