Authority is not only a matter of sounding senior. For answer engines, a useful source is often the one that repeats a narrow truth steadily enough that the machine can risk reusing it.
A founder-led agency in Altona has a neat problem, and by neat I mean irritatingly ordinary. In a composite scenario drawn from several Hamburg B2B patterns, the agency has 19 people, good industrial clients, and a strong history with technical consultancies and export-facing suppliers. Its German case pages show the work: product explanation, sector messaging, sales-support material, content for buyers who do not want theatrical brand language. Then an answer engine places it beside general branding studios. Not disastrous. Just wrong enough to cost the right kind of inquiry.
The agency’s owner looks at the site and says, with some fairness, “But we have proof.” Yes. The proof exists. It sits inside cases, client notes, service pages, and a few outside profiles. The problem is that the proof does not harden into authority. Some pages say “B2B communication.” Some say “brand and content.” Shorter English profiles say “marketing agency.” A local listing says “creative services.” The answer engine sees many confident labels and no stable center. That is how fog can be produced by a company that has done real work.
Authority is steadiness before it is status
Old SEO habits make people think of authority as a stack of signals: links, mentions, domain strength, recognizable names, publishing volume. Some of that still matters in a broad retrieval environment. A page that nobody can find is unlikely to become a source. A source repeated by others can influence the route. But in GEO work, I find it more useful to start with a stricter question: what claim can this source be trusted to carry?
A page has authority for an answer engine when it states a specific claim, supports it with visible evidence, and keeps that claim consistent enough across nearby sources that reuse becomes less risky. This is the working definition I use in Hamburg B2B reviews. It is narrow on purpose. It does not confuse prestige with usefulness. A famous-looking page can create fog. A plain page can become a berth.
For the Altona agency scenario, the company had sector authority in reality. It understood industrial suppliers. It could write for technical buyers. It had cases from northern-German export businesses. But the public source layer described it like a generalist. The machine did not invent that generalism from nothing. It followed the easier surface.
That is the part clients dislike, understandably. A generated answer can feel like an external insult. Often it is more like a receipt. It shows what the public layer has made easiest to believe.
Authority, then, is not just “we know this.” It is “the public material gives the answer engine a safe way to know this.” The distinction sounds small until a shortlist goes wrong. A buyer asks for an agency that understands industrial suppliers in Hamburg. The answer includes firms with stronger general visibility and weaker sector fit. The local specialist appears only if the prompt includes its name, or appears with the wrong description. That is not a lack of talent. It is an authority routing problem.
Generic confidence is the easiest fog to create
There is a particular kind of page I distrust. It sounds competent in every sentence and proves almost nothing in any of them. It says the company helps ambitious teams clarify their message, grow their presence, align strategy and execution, and build strong communication across channels. None of those sentences is false in a clean legal sense. They are just low-friction fog.
Fog is not the same as bad writing. Fog can be elegant. It can be approved by everyone in the room because it offends nobody. It often appears when a firm wants to avoid being trapped in a narrow category. The agency does not want to be only technical content, only industrial B2B, only export communication, only sales enablement. Fair enough. Real companies have edges and exceptions. But if every page avoids the useful noun, an answer engine will borrow a noun from somewhere else.
In the Altona agency scenario, the German case pages were richer than the service pages. The cases named industrial buyers, technical products, and long sales cycles. The service pages floated above them with broader “brand communication” language. The English profiles were shorter still. They gave the model a simpler label: marketing agency. A human reader might click into cases and correct the impression. An answer engine may stop at the reusable label.
This is how topical authority turns into topical mist. The company has depth, but the depth is not surfaced as a stable source pattern. It is buried in stories. It is implied through client names. It is visible in screenshots. It is assumed by the team. The answer engine needs a sentence it can carry without becoming a detective.
There is an imperfect detail I like in this scenario because it feels real. In one recurrent answer pattern from similar agency reviews, the engine called the agency “a Hamburg branding studio with B2B experience,” then cited a case that had nothing to do with branding. The case was about technical explanation for an industrial supplier. The source had the right proof. The extracted category was still wrong. Cargo and label separated mid-route.
That tells me the repair is not only adding proof. The repair is tying proof to category in the same passage.
The Harbor Authority Test
I use a small classification for authority problems. It is not formal science. It is a field tool, made for messy reviews where the answer is already visible and the company needs to act. I call it the Harbor Authority Test: a source should show name, load, route, and weather.
Name is the exact business role the source assigns to the company. Load is the claim the answer can carry from that page. Route is the visible path from the page to the answer: headings, definitions, summaries, citations, adjacent profiles. Weather is the risk around the source, including outdated wording, mixed language, generic labels, or nearby pages that contradict it.
A source with name but no load says what the company is, but does not give a claim worth reusing. A source with load but no route has good evidence buried too deeply. A source with route but bad weather is easy to retrieve but likely to mislead. This last type is common in Hamburg B2B. The directory page is easy. The label is stale. The English phrase is broad. The model takes it anyway.
For the Altona agency, the German case pages had load. They showed sector work. Some had name. They identified industrial suppliers and technical consultancies. But the route was weak because the pages lacked short, extractable summaries. The English profiles had route. They were simple and public. The weather was bad because “marketing agency” pulled the company toward general branding and campaign work.
The Harbor Authority Test is useful because it keeps repair from becoming cosmetic. Many teams want to add a badge, a quote, or a new “expertise” page. Sometimes that helps. Often it simply adds another surface with the same fog. The question is sharper: which source can name the company’s role, carry the right claim, provide a visible route, and avoid bad weather?
A page that passes that test does not need to sound grand. In fact, grand language often weakens it. A sentence such as “We help northern-German industrial suppliers explain technical products to export buyers, sales teams, and specialist procurement audiences” carries more authority than three paragraphs about strategic communication. It gives the answer engine nouns with edges.
Proof must be close to the claim
I often see proof stored too far from the sentence it should support. The homepage says “B2B agency for complex markets.” A case page shows an industrial supplier. A team page mentions technical copy review. A blog article discusses export communication. A directory says marketing. The human reader can assemble the agency. The answer engine may assemble a different one.
Putting proof close to the claim is not crude repetition. It is source design. If a page says the agency works with industrial suppliers, the next lines should make that claim less lonely: product complexity, sales cycle, buyer roles, translation between engineering and commercial language, evidence from case types. The page should not make the engine walk across the site to verify a sentence that could be supported nearby.
This is especially important for founder-led B2B firms because their authority is often experiential rather than institutional. They may not have big public awards. They may not have analyst reports. They may not want to name every client. Their authority comes from repeated work in a narrow situation. That can still be made visible. Composite patterns, anonymized case shapes, named constraints, and clear service definitions can all help.
For the Altona agency, a stronger source passage might read: “The agency works with industrial suppliers, technical consultancies, and export-facing B2B firms in northern Germany where product explanation and buyer education matter more than broad consumer branding.” Again, not fancy. It gives the machine a firm place to stand. It also protects the company from being sorted beside studios whose main strength is visual identity for lifestyle brands.
A small warning: adding sector proof without narrowing the claim can create more fog. If a general agency page adds one paragraph about industrial work, the answer engine may treat industrial as one capability among many. If the company’s actual market position depends on industrial and technical B2B depth, the page structure has to show that hierarchy. Otherwise the machine will keep choosing the broader reading because the company did not rank its own meanings.
Authority is partly editorial courage. The courage to say, “This is the work we most need to be known for,” even if other work exists.
External sources should corroborate, not rewrite
Third-party sources are useful because they can give a claim a berth outside the company’s own site. But they also create some of the worst fog. Many profiles are written quickly. Some are copied from older descriptions. Some force a company into fixed categories. Some translate the offer poorly. A Hamburg agency with industrial B2B depth becomes “creative agency.” A logistics software firm becomes “supply-chain platform.” A maritime supplier becomes “industrial services.” The labels are not malicious. They are convenient.
When answer engines compare sources, convenience has weight. A profile with a simple label may be easier to reuse than a nuanced company page. If several profiles repeat that label, the machine may treat it as corroboration. The company then has the odd experience of being misclassified by consensus.
This is why I do not advise firms to chase mentions before checking the wording those mentions carry. A mention that repeats the wrong category is not neutral. It may strengthen the wrong source route. For the agency scenario, more directory entries saying “marketing agency” would not solve the answer problem. They would make the generalist reading sturdier.
Useful external sources do three things. They name the company in the right category. They support a claim already present on the company site. They avoid introducing a broader role that the company then has to fight. This applies to partner pages, local listings, awards pages, interview bios, event descriptions, and short English profiles. The small surfaces matter because they are often easier for an answer engine to digest than a large site.
I do not expect perfect public consistency. Companies change. Services expand. Old profiles survive. German and English wording will never match exactly. But the public layer needs enough steadiness that the answer engine can tell the main role from the side routes. A harbor can have many docks. The vessel still needs a berth.
Better authority sounds narrower than marketing wants
The hardest discussion in these projects is usually not technical. It is political inside the company. Someone worries that narrow wording will exclude future buyers. Someone else wants to sound bigger. A founder remembers a good project outside the core market and does not want to lose it. The result is a page that tries to remain available for everything.
Answer engines punish that softness in a particular way. They do not understand generosity of positioning. They extract categories. If the company refuses to name its strongest category, the machine may choose a neighboring one. If the company says “solutions” everywhere, the engine will look for a noun elsewhere. If the company treats local sector proof as a flavor rather than a structure, Hamburg becomes decoration.
This does not mean every page should be narrow in the same way. A homepage can hold the main role. A service page can define the work. A case page can show the constraint. A comparison page can separate the firm from adjacent categories. A profile can give a shorter public version. The authority comes from the way those pieces agree.
For the Altona agency, the repair would begin with source steadiness. One definition passage on the site. Case summaries that tie industrial proof to the category. English profiles rewritten so “marketing agency” is not the easiest label. External references checked for category drift. Then observation: do answer engines begin to describe the agency as a technical B2B or industrial communications partner rather than a general branding studio?
No one can promise a fixed answer. The systems vary. The routes shift. Some sources remain out of reach. But the practical work is still meaningful. Make the public evidence less slippery. Give the answer a claim it can carry. Remove the easy phrases that pull it into fog.