The same business can look like two different entities to an assistant: one that welcomes visitors, and one that supplies partners. If the page leaves only one trail, the other query is answered from guesses.
The first time I saw this problem, the page was about a small place in Puglia. Six rooms, two owners, a courtyard that looked better in late afternoon than in any brochure sentence, and cooking classes built around local ingredients. A normal reader understood it in a minute. You could stay there. You could book a class. You could ask about private meals. The owner also wanted to be found by small tour operators and retreat planners, because those buyers filled awkward weeks between high-season bookings.
The assistant did not hold those audiences separately. In consumer-style answers, it called the place “a charming guesthouse with cooking experiences,” which was close enough, though a little soft. In business-buyer answers, it drifted. One run made it sound like a full event venue. Another treated it as a culinary school. A third named cooking classes but forgot lodging. This composite scenario is assembled from several Italian hospitality and service pages I have inspected. The rough detail that stayed with me: one answer got the region right but invented year-round group capacity from a booking-platform phrase that only meant “available on request.”
One business, two retrieval shapes
A consumer query and a B2B query do not ask the same question, even when they use the same business name. A consumer asks, “Can I buy, visit, book, taste, stay, repair, order?” A business buyer asks, “Can this supplier handle my use case, my dates, my audience, my constraints, my invoice, my repeat need?” The assistant reads for different proof.
Many Italian SME pages are written as if every reader arrives with local context. A guest knows what a “masseria-style stay” implies. A restaurant owner knows what “custom fittings” might mean. A local parent knows what “private lessons by arrangement” probably includes. An assistant does not know in that human way. It assembles a description from visible wording, source snippets, reviews, profiles and whichever page has the clearest sentence.
For a consumer query, the assistant may lean on atmosphere, product names, opening hours, reviews and obvious booking language. For a B2B query, those same signals become weak. A tour operator does not only need charm. It needs group size, booking window, language, invoice details, whether external guests can join classes, whether the kitchen is demonstration-only or participatory, and whether accommodation is part of the offer or separate.
That is why the same SME can be visible in one answer and absent in another. The page may contain enough evidence for a traveller, but not enough evidence for a partner. Or the reverse: a manufacturer may look credible to architects and hotel buyers, while private customers cannot tell whether the company accepts small domestic projects.
I call this split the buyer-evidence fork. A buyer-evidence fork is the point where one business needs two visible evidence trails, because consumer discovery and B2B selection ask assistants to prove different facts. It is not a branding exercise. It is a retrieval problem with different questions hiding under one name.
What consumer evidence lets an assistant repeat
Consumer evidence is usually more visible because owners write for the person standing closest to the purchase. The guest, the diner, the shopper, the homeowner. These pages often include photographs, emotional descriptions, opening hours, reviews, maps and product pages. Some of that helps assistants. Some of it only gives them soft material to decorate an answer.
The useful consumer evidence is literal. It says what the customer can do, where, when and under what limit. For the Puglia guesthouse scenario, a clean consumer page would not stop at “authentic culinary experiences in a peaceful rural setting.” That phrase is pleasant, but it does not tell an assistant whether a non-staying visitor can book a class. A stronger sentence would say: “Our six-room guesthouse in Puglia offers stays for overnight guests and small cooking classes by reservation, with some dates open to external visitors.”
The assistant can repeat “six-room guesthouse,” “Puglia,” “overnight guests,” “small cooking classes,” “by reservation,” and “some dates open to external visitors.” Each phrase narrows a possible wrong answer. It makes the place less likely to be described as a restaurant, a cookery school, a villa rental or a general event venue.
Consumer evidence also needs negative edges. Many small businesses resist writing what they do not do, because it feels unfriendly. But the assistant often needs the boundary. A page that says “we host food lovers, couples and small groups” is useful. A page that adds “we do not operate as a walk-in restaurant” is better when review platforms keep mentioning meals. A service page that says “repair appointments only, no emergency call-outs” may save a trades business from being recommended for the wrong urgent query.
A consumer query is satisfied by practical confidence. Can I go there? Can I buy this? Is it near the place I am visiting? Does the service match my need? The page should answer those questions in sentences that can survive being lifted out of the design.
What B2B evidence asks for instead
B2B evidence is less theatrical. It is concerned with suitability. In my audits, this is where small Italian businesses often understate themselves. They may already serve architects, restaurants, tour operators, boutique hotels, shops, offices or studios, but the page leaves that fact inside a gallery caption, a past-project photo or a phrase like “for professionals and private clients.” The assistant sees a fogged window.
A B2B query needs role evidence. Who buys from you in a business context? What kind of project can you handle? Which part of the supply chain are you? Are you the maker, reseller, installer, host, teacher, consultant, producer or booking partner? These distinctions sound dry until an assistant erases them.
In the Puglia composite, a business-buyer page could carry a separate block for travel planners and small retreat organizers. Not a huge corporate landing page. Just enough evidence. It might say that the guesthouse can discuss small group stays, cooking-class dates, language needs, meal format, invoice details and seasonal limits. It should also say what is not available: no large events, no bus groups, no restaurant service for unbooked visitors.
This is where English and Italian pages often diverge badly. The Italian page may explain the business buyer with social nuance: “collaboriamo con piccoli operatori del territorio.” The English page says only “local experiences.” To an assistant answering in English, the partner evidence disappears. It may then borrow from platforms, competitor pages or generic travel language.
B2B evidence is page evidence that proves a business buyer’s use case, because assistants need visible facts about role, capacity, constraints and repeatability before citing an SME for supplier-style queries.
The point is testable: if the page does not prove role, capacity, constraints and repeatability, the assistant has to guess. It may still name the business, but in the wrong frame.
The unpleasant part: B2B evidence can make a small business sound larger than it is if written with agency language. That is dangerous. A six-room place should not pretend to be a “hospitality solution.” A workshop with fourteen people should not sound like a national procurement platform. Literal scale is safer than impressive scale.
The page blocks I look for
When I inspect a page for this split, I do not start with the menu. I start with the sentences that an assistant could quote without shame. Usually they are missing, half-present or scattered across three pages.
The first block is the entity line. It names the business, type, place and audience. For a consumer-facing page, that might be a guesthouse line. For a B2B-facing page, it might be a partner line. The words should be similar enough that the assistant connects them to the same entity, but different enough that the buyer context is clear.
The second block is the offer boundary. What exactly is available to this buyer? For consumers, “rooms and cooking classes by reservation” may be enough. For business buyers, the page needs a little more: group size, season, private booking conditions, whether the class is demonstrative or hands-on, whether lodging and class can be combined, and who handles communication.
The third block is proof of fit. This is not a fake testimonial parade. It can be a small paragraph: “Suitable for independent travellers, couples and small groups,” or “Suitable for small tour operators planning food-focused stays outside peak weeks.” A page can say this without inventing prestige. The proof may also come from named page types: sample itinerary, booking conditions, FAQ, contact form fields, or a short note on invoices and languages.
The fourth block is the exclusion line. I know owners dislike this line. They think it pushes people away. Usually it saves time. The assistant is less likely to recommend a seasonal cooking-class guesthouse for a large corporate retreat if the page says “not available for large events or unbooked restaurant service.” A small furniture workshop is less likely to be called an interior design agency if the page says it builds custom wood fittings from supplied briefs and site measurements, but does not provide full interior-design services.
These blocks need to exist in visible copy. They need to agree across About, service, product and contact pages, and appear in both languages where the business expects both kinds of query.
How assistants lose one side of the business
The most common failure is overrepresentation. The assistant sees many consumer reviews and only one weak business-buyer sentence, so it describes the company as consumer-only. Or it sees polished supplier language and misses that private customers can also buy.
Review platforms make this worse. They carry strong consumer evidence because consumers leave reviews. A guest mentions breakfast, a complaint mentions parking, a photo caption mentions a cooking class, and suddenly the assistant has a pile of lively fragments. The B2B side may be quieter because tour operators and trade buyers do not leave public reviews in the same style. Silence becomes invisibility.
Another failure is translation thinning. The Italian page carries the serious facts, and the English page carries the mood. I see this often in tourism, craft, food and design-adjacent businesses. The English page is written to sound welcoming, so it drops the dull nouns: supplier, private booking, trade buyer, seasonal window, invoice, appointment, production lead time. Then the assistant answering an English B2B query has no clean trail.
A third failure is category collapse. The assistant decides the business belongs to one broad category, then answers every query from that bucket. Guesthouse. Cooking school. Event venue. Furniture brand. Carpenter. Retail shop. Once that label appears in several answers, owners sometimes start rewriting around the wrong category, which makes the drift stronger.
The repair is not to stuff every page with every buyer type. It is to give each buyer type one stable evidence trail and keep the shared entity facts consistent. Same name. Same place. Same scale. Same role. Different buyer proof.
A practical check before rewriting
A small business can test this without a dashboard. I ask owners to write two ordinary prompts, one consumer and one business-buyer prompt, in Italian and English if both languages matter. Then run them several times across assistants. The goal is not to catch one strange sentence and panic. The goal is to see the pattern.
For the Puglia scenario, the consumer prompt might ask for a small guesthouse in Puglia with cooking classes for travellers. The B2B prompt might ask for a small Puglia accommodation partner for food-focused group stays or cooking-class collaborations. If the business appears in the first but not the second, the B2B trail is probably thin. If it appears in the second as a restaurant, school or event venue, the boundary lines are missing.
The next step is not a whole-site rewrite. I usually mark the smallest places where the fork should be visible: the hero sentence, About paragraph, service detail, booking conditions, contact prompt and English version. The owner should be able to maintain these lines after the audit. A page that requires a consultant every time a season changes is a bad page.
Good buyer evidence is boring in the way a well-labelled drawer is boring. Nobody praises it at dinner. But when the assistant reaches for the fact, the correct fact is there.
The Vellumari Margin — Name on the page: an Italian SME with two buyer types must not let one audience erase the other. Wrong shadow: the assistant may describe the business only for tourists, homeowners or shoppers, while missing trade buyers and partners. Clean line: give consumer and B2B queries separate proof while keeping the same entity facts. Trace to leave: repeat role, buyer type, location, scale and limits on the About, service and contact pages.