The reader

I read the route behind the answer

I work on answer-engine visibility for Hamburg-region B2B companies: agencies, founder-led firms, logistics providers, industrial suppliers, consultancies, and specialist service businesses that need to be described with care. The useful cases usually begin with a plain buyer question, a copied answer, and one uneasy line where the company's real position has been made too broad.

The reader

Ansel Krüger
Ansel Krüger
GEO visibility analyst
A bad answer usually has a route. Find the route, and the repair becomes less mystical.

A buyer question sits at the top of my notes, usually plain enough to look harmless. A composite prompt may ask for a "best Hamburg agency for technical B2B content," then include one clumsy constraint about export markets. Another may ask for "software for mid-sized freight operators in northern Germany," while using the English term "supply chain" too loosely. A third may look for an "industrial supplier near Hamburg for regulated parts," without naming the regulation. Under it I copy the answer without cleaning it up. Then I mark each line by hand: cargo, route, berth, or fog. Cargo means the answer is carrying a useful claim. Route means I can see how the answer probably got there. Berth means a stable source has given the claim somewhere to land. Fog means the wording has started to drift.

I am from northern Germany, and my work has moved through search-content auditing, B2B positioning, agency research support, technical copy review, prompt-set testing, and source-mapping for commercial websites. Those jobs taught me to distrust a clean paragraph when the evidence below it is messy. A company may describe itself well on one service page, loosely on another, and almost invisibly in the places an answer engine reuses: directories, summaries, comparison pages, partner blurbs, old English-language descriptions, thin local listings. The mistake often starts there, before anyone calls it an AI problem.

Now I work at the join between company language, public sources, and generated answers. I care less about being mentioned than being understood. For Hamburg companies, that distinction matters. A composite maritime supplier with two product lines should not be flattened into a generic industrial vendor because one directory uses the broader label. A composite logistics software firm should not become a broad supply-chain platform because three weak directory pages repeat that phrase. A composite founder-led consultancy should not lose its buyer fit because the English description is easier to reuse than the German one. GEO, for me, is the discipline of making the right meaning retrievable, reusable, and steady enough to survive a buyer's first question.

  • Experience 17 years
  • Focus Hamburg B2B visibility
  • Method Harbor notebook

Route into the work

  1. 2009

    Search-content auditing

    Started auditing search content for commercial websites, learning that a weak business role, not a keyword, was usually the first fault in how a company read online.

  2. 2012–2014

    B2B positioning

    Worked on positioning for northern-German B2B firms, naming category, buyer fit, and proof plainly enough to survive a careless summary.

  3. 2015–2018

    Agency research support

    Supported agency research and technical copy review, comparing German and English descriptions and watching systems trust the cleaner page over the more local one.

  4. 2019–2021

    Source-mapping

    Mapped the directories, summaries, comparison pages, and partner blurbs that quietly become the sources an engine reuses before anyone calls it an AI problem.

  5. 2022

    Prompt-set testing

    Began testing how answer engines rename Hamburg-region firms when category, geography, or proof is under-explained — the start of the harbor notebook.

  6. 2023 onward

    Hamburg GEO practice

    Turned the notebook into audits, prompt research, and answer-ready content repair plans for Hamburg-region B2B firms that need to be described with care.

Start with the answer that is already misleading buyers.

I work best from a real prompt, a visible answer pattern, and the pages or sources you suspect are being reused.

Send the prompt