--- title: "artheris · Become the brand AI recommends." description: "We measure how your brand shows up in AI answers, and change it with evidence: generative engine optimization (GEO). Open alpha, limited seats." canonical: "https://artheris.de/en" lang: "en" alternate: "https://artheris.de/" describedby: "/llms.txt" --- [HTML version](https://artheris.de/en) # artheris · Become the brand AI recommends. ## Does AI recommend your brand? Ask what your customers ask. Artheris shows who gets recommended, why, and which change you can roll out under your control. [Interactive version](https://artheris.de/en#composer) ## Measurement is solved. Auditable action is not. *What you get* The market splits into platforms that see a lot and change nothing, and automations that change a lot and document nothing. We do both. Software carries out the execution, you keep the decision. Terms - LLM journey: A buying process in which the buyer questions an AI system before building a shortlist. - Recommendation gap: Share of purchase-close queries in which your own brand is not recommended. - Opportunity rate: Share of journeys that would become a qualified sales opportunity without the gap. - Win rate: Share of sales opportunities that sales ultimately wins. - Organization schema: Machine-readable company data in the source code that AI systems read. - Draft pull request: A prepared code change in the repository that only merges after approval. Example calculation for a B2B SaaS company, not measured values: every number is an adjustable assumption. This is what it looks like in the product. - Step 03 · Proposal - Ready to approve, not a suggestion - You do not get a to-do list. You get the finished change, with current and target state. Example: proposal v1 · organization schema derived from finding 2, waiting for approval - Step 04 and 05 · Approval - Approval before every change - Nothing goes live without a person. No auto-publish, not even as an option. A rejection is logged exactly like an approval. Example: audit record A-1042 every decision writes a record, a rejection too - Step 06 · Verification run - Remeasurement under unchanged methodology - The result comes from a new run under unchanged methodology, not a conveniently chosen comparison. Example: remeasurement · 3 systems · catalog v3 same questions, same counting, same catalog. A deviating run drops out of the comparison. Example views. The values are examples, the format is the product. ## 01 · Measure the market We measure what your buyers actually see. Artheris runs purchase-relevant questions repeatably across relevant AI systems and captures recommendations, sources, competitors and answer context as a solid baseline. Output: Comparable answers with model, market, source and timestamp ## 02 · Evidence the causes Answers become traceable causes. Bots and an analysis layer connect prompts, statements, citations and source pages. Measured facts stay separate from hypotheses, so every recommendation stays checkable. Output: Evidence map with source path and a reasoned opportunity ## 03 · Prioritize opportunities The team works the strongest lever first. Artheris scores every opportunity by expected impact, strength of evidence, effort and risk. The result is a clear order, not a long backlog. Output: Prioritized actions with effort, priority tier and delivery path ## 04 · Prepare the change Every recommendation becomes a concrete change. The orchestration layer translates the finding into a checkable content or technical diff. CMS, PR and edge lanes prepare execution without publishing on their own. Output: Implementation-ready diff with rationale and target system ## 05 · Release under control A human keeps the final decision. Changes move through defined approvals and only ship over the connected lane after sign-off. Version, owner and rollback path stay in the audit trail. Output: Approved release with audit trail and rollback ## 06 · Prove the impact After the change, verification begins. Artheris repeats the baseline run under comparable conditions, measures the change and feeds new market shifts back into the next controlled optimization cycle. Output: Before-and-after evidence and the next prioritized loop ## Six steps, one architecture. *The surface behind the six steps* Signals come in at the front. Agents prepare, people decide, and an implemented change with a record leaves at the back. Several people and agents work on that same surface at the same time. Schematic · one pass - Checked without an AI verdict: Deterministic on structure and schema - Change queue: Proposal v1 is waiting, with current and target - People decide: Approval or rejection, one record each - Lanes: Draft pull request · CMS · edge Schematic, not a product screenshot. The chambers are steps 02 to 05, the re-measurement closes the loop back to step 01. Further tracks run through the same connection, the same approval and the same chain of evidence. GEO, generative engine optimization, is the first module; the last slot is open. ## Frequently asked questions The questions we hear most often, with the answers we can back up. ### Is this not just SEO? No, but it does not replace it either. SEO gets your page into a list of results. This is about whether your brand appears in an answer that an AI system writes itself. Different place, different lever. The term for it is GEO, generative engine optimization. ### AI answers vary. So what are you actually measuring? They do vary. An independent study by SparkToro covering 2,961 queries found the same list of brands in under one percent of cases when the same question was asked again. We do not claim otherwise, and we do not report a ranking position within an answer, because there is none. We put a fixed set of questions to several systems and count how many answers mention your brand. That is AI share of voice: mentions of your brand divided by all mentions in the comparison set. Only answers that triggered an actual web search are counted. A single measurement is a sample, not an exact figure, and we tell you how many questions it is based on. ### Could I not just check this myself by asking ChatGPT? For a first impression, yes, and we recommend it. Open a fresh chat and ask for providers in your field. What you do not get that way: the same set of questions across several systems, the same counting method over months, and a record you can show to someone. ### Will this bring revenue, or is it a vanity metric? Visibility in AI answers is not traffic, and we do not promise you any. It is the point where a shortlist forms before anyone clicks at all. We show you that shortlist and how it changes. Deriving a revenue effect from a movement inside a measurement window would be an attribution, and we do not make those. ### What happens when a provider changes its model? The answers change, and you see it in the measurement. These systems are not ours, we do not know their changes in advance, and nobody can guarantee you anything there. What we do not do is smooth over the break. Runs whose methodology does not match drop out of the comparison rather than producing a line that does not exist. ### Where is our data held? In the EU. The waitlist database is located in Frankfurt am Main. For this page we store only what you enter in the form: your email address and, if you want, your domain. Everything else, including data processing under Art. 28 GDPR, we settle in writing before a pilot starts. ### Is this finished, or still alpha? It is alpha, and we call it that. The open alpha takes on a limited number of companies. You get early access and help turn the product into a reliable workflow. If you need a finished product today, you are too early for us. ## We take companies on through the waitlist. [Request access](https://artheris.de/en#access) Open alpha. No newsletter, no data sharing. Confirmation by email.