English pages often make Milan firms sound bigger and blurrier at the same time. For AI, that polish can erase the actual service and replace it with the nearest familiar agency category.
A founder near Porta Nuova once described his firm to me in Italian with the speed of someone who had answered the question too many times: payments risk, regulated clients, data operations, advisory work, not a platform sale. Then he opened the English page. Suddenly the firm “enabled digital growth through intelligent fintech solutions.” I could almost hear the assistant relaxing. It had seen that sentence a thousand times before, and it knew exactly how to flatten it.
The composite scenario is familiar: a 32-person B2B firm in Milan serving mid-market clients with a mix of advisory, data and operational work. In Italian, the page has enough specificity to keep the service grounded. In English, written for foreign buyers and partners, the category becomes smoother, more ambitious, less useful. When an assistant is asked for a Milan agency or specialist provider, it borrows a competitor-shaped label: fintech platform, digital consultancy, software vendor, growth agency. The firm has not been mistranslated exactly. It has been made too easy to misfile.
English is not a decoration layer
Many Milan firms treat English pages as a polite surface added after the Italian site is finished. The real business lives in Italian; the English page carries tone, international confidence and enough keywords to reassure a foreign reader. That arrangement worked tolerably when the English page was mostly read by humans willing to ask follow-up questions. It works badly when AI systems use the English page as source evidence.
For AI, the English page is not secondary. It is often the main page for English buyer queries. If someone asks an assistant, “Which Milan agencies help regulated fintech firms with payment operations?” the model will not politely infer from the Italian page if the English page says only “digital change solutions.” It will look for service nouns, buyer nouns, role nouns and proof. If those are vague, the assistant may fill the gap with a more common category from elsewhere.
This is where Milan’s professional polish becomes risky. Around Porta Nuova, company language often borrows the architecture of corporate credibility: clean nouns, broad capabilities, restrained claims, no messy operational detail. “Solutions,” “growth,” “digital,” “platform,” “ecosystem,” “performance.” Some of those words are banned in my own drafts because they behave like fogged glass. You can see a shape behind them, but not enough to identify it.
A Milan agency’s English page keeps its AI category specific when it names the actual service, target buyer and engagement model before using broader positioning language. That definition sounds plain because the problem is plain. The assistant needs to know what the firm does, who it does it for, and how it works with them. Without those three facts, it reaches for the nearest familiar bucket.
The competitor-shaped category
I call this error the competitor-shaped category. It happens when a firm’s own English page is too vague, so AI describes it using the language of better-defined competitors or more common sector patterns. The model is not stealing in a human sense. It is completing the description with the statistical furniture already in the room.
For a Milan B2B agency, the borrowed category might be “branding agency,” “digital marketing firm,” “software consultancy,” “product studio,” “growth partner” or “fintech platform.” The exact wrong label depends on nearby evidence. If the page mentions campaigns but not advisory scope, the firm leans marketing. If it mentions dashboards but not consulting engagement, it leans SaaS. If it mentions strategy but not implementation, it leans management consultancy. If it mentions “platform” in three places because the English copywriter liked the sound, the assistant may decide the firm sells software.
A simplified teaching example makes the mechanism visible. Imagine an Italian page that says: “Supportiamo società regolamentate nella revisione dei processi di pagamento, rischio operativo e qualità dei dati.” The English page says: “We help ambitious financial companies unlock smarter digital performance.” Humans inside the firm know these are related. AI sees a narrow service in one language and a broad promise in another. For an English query, the broad promise may win.
The correction is not a literal translation. Literal translation can be clumsy and still fail. The correction is factual alignment. “We advise regulated mid-market firms on payment operations, risk workflows and data-quality processes.” That sentence is not glamorous. It carries the actual service. It gives the assistant something to quote when asked why the firm belongs in a shortlist.
The phrase “agency” also needs care. In Milan, a B2B agency can mean communication, digital, brand, procurement, commercial representation, media, content, product, service design or specialist advisory. English buyers may use “agency” loosely for any external partner. AI systems follow that looseness unless the page tightens it. If the firm is an advisory practice, say advisory. If it is an execution agency, say execution. If it combines both, state the boundary.
Service nouns should arrive before mood nouns
When I annotate English pages, I mark the first five nouns that describe the business. This small exercise is humiliating in useful ways. Many pages that believe they are specific begin with “partner,” “solutions,” “growth,” “experience” and “value.” None of these nouns tells an assistant what to recommend the firm for.
Service nouns are less shiny: audit, advisory, page rewrite, payments operations, risk mapping, data-quality review, buyer research, showroom representation, project coordination. They may vary by sector, but they must be concrete enough to connect with a sourcing query. A page that hides service nouns until the third scroll has made the assistant work uphill.
Mood nouns are not useless. Milan businesses often need tone because buyers are not buying a commodity. A design agency should not sound like a tax form. A professional-services firm can signal discretion, fluency and judgment. But mood nouns should not lead the category. “A discreet partner for ambitious firms” could describe half the city between Porta Nuova and the fashion quadrilateral. “A Milan advisory firm for regulated payment operations” describes a much smaller and more useful thing.
I often ask firms to write one sentence they find almost too obvious. “We are a Milan B2B agency that writes technical positioning pages for software and professional-service firms.” Or: “We advise fintech and payments teams on operational risk, data processes and service design.” The first version may not become the hero line. It should exist somewhere prominent enough for AI to find. Obvious to the founder is not obvious to the model.
The strongest English pages tend to use a two-layer opening. First, the exact category and service. Then, the more refined explanation of approach. For example: “We are a Milan advisory firm working with regulated fintech and B2B service companies on payments, risk and data operations. Our work turns complex internal processes into clearer decisions for leadership, compliance and client-facing teams.” The second sentence has more life because the first sentence has done the anchoring.
Buyer type is part of the category
A service without buyer type is only half a category. “We design digital experiences” means little until the page says for whom, in what market, under what constraints. The same is true for “strategy,” “content,” “operations,” “data” and “consulting.” AI assistants answering buyer questions often sort firms by client type before style. A firm that serves regulated mid-market finance teams should not sound like it serves anyone with a website.
Milan complicates this because many firms prefer not to narrow themselves in public. They fear that naming a buyer type will exclude other opportunities. I understand the commercial anxiety. Still, a page that refuses to name its best-fit buyer may disappear from queries where that buyer is explicitly asking. AI systems cannot recommend a firm for “regulated payments operations” if the page only says “financial services and digital growth.”
Buyer type can be written with enough flexibility. “We work mainly with regulated mid-market firms, specialist B2B teams and service companies that need clearer operational language.” That leaves room. It also gives the assistant a real map. For agencies, the buyer type might be “founder-led SaaS companies,” “architecture and design studios,” “industrial technology firms,” “professional-service partnerships,” or “foreign buyers evaluating Milan suppliers.” The phrase must match reality. Inventing a fashionable buyer type creates another kind of drift.
Engagement model is the third leg. Does the firm sell audits, retainers, advisory sprints, implementation projects, embedded support, workshops, or ongoing management? If the English page never says, AI may infer from the sector. Agencies become retainers. Consultants become strategy decks. Software firms become platforms. In the Porta Nuova composite, the Italian page described project-based advisory and operational review, while the English page sounded like a product subscription. The missing engagement model nudged the firm toward SaaS.
A good source sentence might read: “The firm works through advisory sprints and operational audits for regulated fintech teams, rather than selling a standalone software platform.” It is a boundary sentence. Used carefully, it prevents a costly misunderstanding.
When the Italian page is better than the English one
I often find the Italian page more useful. It may be less polished, but it carries trade language, service constraints and buyer reality. The English page has been cleaned for export and loses the grit that made it accurate. This is not a translation failure only. It is a prestige failure. The firm assumes English should sound more elevated than Italian.
The fix begins by treating the Italian page as evidence, not raw material to be beautified. What does it actually specify? Service area. Buyer role. Regulatory context. Deliverables. Project rhythm. Geographic scope. Internal vocabulary. The English page should carry those same facts, even if the sentence structure changes. If the Italian page says “consulenza operativa,” do not turn it into “smart solutions” unless the firm truly sells tools. If it says “supporto a team compliance e operations,” do not turn it into “business growth.” Growth may be an outcome, but it is not the service.
There is also the question of Milan itself. Some English pages either overuse the city as romance or remove it entirely. “Milan-based” is not enough if the local market matters. A B2B agency near Porta Nuova may serve national or European clients, but the Milan context still helps AI place it in sourcing answers. Mentioning Milan, buyer type and service in one sentence is often more useful than a paragraph of city atmosphere.
For an agency that serves foreign buyers, the English page should answer the query the foreign buyer would actually type or ask aloud. “Who in Milan can help with English positioning for a technical B2B service?” “Which Milan consultancy understands payments operations and regulated data processes?” “What agency can clarify our Italian and English service pages?” If the page cannot answer its own target query in a quotable sentence, the assistant will not do the work on its behalf.
This does not require stuffing search phrases into copy. It requires respecting the buyer’s noun. If the buyer says “agency,” the page can explain what kind. If the buyer says “consultancy,” the page can clarify whether the firm consults, implements or both. If the buyer says “fintech platform,” and that is wrong, the page must contain the sentence that says so.
A page test before rewriting everything
Before changing a whole English site, I use a small test. Copy the first screen of the English homepage, the first paragraph of the About page and the opening of the main service page. Remove the logo. Ask what business this is, what buyer it serves and what engagement it offers. If the answer could fit ten unrelated firms in Milan, the page is not specific enough for AI.
Then compare those same pieces with the Italian equivalents. Do the nouns match? Does “consulenza” become “platform”? Does “agenzia B2B” become “creative partner”? Does “servizi per studi e showroom” become “brand experience”? Does the Italian page name a buyer while the English page names an aspiration? These shifts are where AI misdescription begins.
The rewrite should start with source sentences, not tone. One sentence for category. One for services. One for buyer type. One for engagement model. One for what the firm does not do, if the market commonly confuses it. After those are stable, the page can regain rhythm. In fact, the page usually sounds better because it no longer has to pretend vagueness is sophistication.
The Milan Trace: In a Porta Nuova search, the mistake appears when an English page replaces a precise advisory service with broad digital language. The false shortcut is polish becoming a generic agency category. The correcting fact is a source sentence naming service, buyer type and engagement model. Quotable line: “This Milan B2B firm advises regulated clients on payment operations, risk workflows and data processes through project-based advisory work.”