AI Visibility: What AI Assistants Say About Your Company

Because a language model does not reproduce your website - it reproduces the sum of what it finds about you: directories, old press releases, review portals, competitor pages, archived versions of your own pages. If your current presentation is not the strongest source, the outdated one wins. Obsolete prices, discontinued products and wrong contacts almost never come from a hallucination; they come from a real, outdated source.

• Directory entries nobody has maintained for years
• Press releases about products that no longer exist
• Competitor pages and comparison portals that describe you too
• Your own website, technically unreadable for AI crawlers

A score tells you that you are at 34 out of 100. It does not tell you why, and it does not tell you what to do on Monday. ArkeonTech delivers the source list instead: which specific pages the systems draw on, which claim comes from which source, and which of them can be influenced. You can check the technical part yourself beforehand - free, no signup, with the AI visibility check.

We jointly define the questions your customers actually ask AI assistants and put them repeatedly to ChatGPT, Perplexity, Google Gemini and Microsoft Copilot. We log not only whether you are mentioned but what is said and which sources the system cites. The repetitions reveal a pattern: which domains define your market in the eyes of these systems, and where you appear in it - or do not.

• Question set built from real customer questions, not keyword tools
• Repeated queries across several systems, because one answer proves nothing
• A log of the sources cited per answer
• Analysis of which domains dominate the market and why

Yes, and it is the core of the analysis. You name up to three competitors and they run through the same question set. The result is a table: how often each is mentioned for which question, in which position, with which claim and from which source. That is the evidence a management team wants before a budget decision, and it is reproducible at the next measurement.

Actions follow from the findings and are ranked by impact. Technical issues on your website we fix directly or hand to your agency as concrete instructions. For incorrect third-party sources we take over the correction requests. For content gaps we produce material the systems can cite. On request we repeat the measurement quarterly so progress stays provable.

• Make your own website technically discoverable
• Correction requests to directories, portals and press pages
• Set up machine-readable company data (Schema.org, llms.txt)
• Citable content for the questions where you are missing
• Quarterly re-measurement against the baseline

Companies with 5 to 250 employees about which enough exists online for a language model to have something to report: providers of explanation-heavy services, manufacturers selling through specialist retail, law firms, medical practices and consultancies. Two triggers account for almost every enquiry - a prospect quotes a false claim from an AI answer, or a competitor gets named and you do not. The analysis runs fully remote and is therefore location-independent: we serve clients in Germany, Austria and Switzerland, with on-site meetings possible around Aalen and Stuttgart.

• Company size: 5 to 250 employees
• Prerequisite: your own website plus mentions in third-party sources
• Benchmark against up to three competitors you name
• Delivered remotely across Germany, Austria and Switzerland

When you expect a commitment that ChatGPT will mention you from now on. There is none, and anyone who offers it is selling something other than an analysis: the sources can be influenced, the answer cannot. Nor does the analysis help against a justified bad review - we correct false claims at the source, we do not make accurate criticism disappear. For a company with no meaningful traces online the measurement is simply too early; what is missing there is not a correction but a foundation, and the first step is a website AI systems can read. And if all you need is a number for reporting, a score tool is faster and cheaper than a source list.

Michael Kaiser, owner of ArkeonTech: business informatics (DHBW Heidenheim), B.Sc. Business Analytics (Aalen University), previously several years as an ERP consultant to mid-sized companies in Baden-Württemberg. He compiles the question set with you, runs the queries and owns the evaluation. The methodology comes from our own operation: ArkeonTech continuously analyses the Search Console and Bing data of its own domain for machine-shaped queries and publishes the findings on the blog.

Email: m.kaiser@arkeontech.de