Your customers now ask the AI. Does the AI know your brand?

Generative Engine Optimization (GEO) is the work of making sure a brand appears in the answers of AI systems such as ChatGPT, Gemini or Perplexity and is portrayed correctly, just as SEO makes sure it is found on Google. augenmerk measures the current state, evidences it and derives the next step.

A single score cannot tell whether the AI does not know a brand or cannot find it in its sources. augenmerk separates model knowledge from live search, measures repeatedly rather than once, reports the noise band, calculates product lines separately and evidences every statement with the passage from the original answer.

No. augenmerk puts the statement before the number: every view answers one question in a sentence, the numbers and evidence sit underneath, and the next steps are written as a file with reason, action and the metric that shows success. The documentation explains the terms, and the AI assistant answers questions about your data right in the application. That makes augenmerk a tool for marketing and PR teams, for management and for shareholders who want to know how the brand stands in AI answers without needing a specialist team. If you have an SEO team, hand them the gap view and the content ideas as a work list.

Ads in ChatGPT appear below the answer when the conversation matches the offer. augenmerk turns your measurements into a campaign plan: ad groups and context hints for the questions where ChatGPT does not mention your brand, and factual answers where the AI discusses you critically. We create the plan in your OpenAI ads account, paused. You review and switch it on. Budget and billing run directly with OpenAI. The feature is in beta, included in Pro and available in Starter for €29 a month; opt in when you register and we enable it for your account.

An assistant inside the application that knows your project. It explains every view, answers questions such as “Why does the model-knowledge channel show no sources?” or “Which page should I revise first?” and points to the evidence behind them. It reads the same data as the views and the public documentation, nothing beyond that, and it does not guess: what it cannot derive from your measurements, it says so. Requests count against your compute units. Starter includes up to 50 requests a month, Pro up to 150.

For marketing and PR teams, for management and shareholders who want to know how their brand stands in AI answers, and for agencies that have to prove it for their clients. German-speaking markets first; the application is available in English.

ChatGPT, Claude, Gemini and Grok each with and without web search, Perplexity and Google AI Overviews: ten channels in three channel types, included in every plan. Google AI Mode follows as soon as it can be measured reliably.

One run a week with all active questions on all channels. The same question gets two different answers on two days; what is reliable is presence across several runs, not a single measurement. A daily full run costs seven times as much and, within the noise band of a question set, adds no insight, just randomness with a date.

The blur between two runs with the same questions, depending on the number of counting questions: around ±10 percentage points with eight questions, around ±5 with 32. Movements below that are not reported as movement. The calculation is in the documentation.

A second model reads from every answer which brands are mentioned and what is said about them, and has to deliver the verbatim passage for it. We check every passage character by character against the stored answer. What is not found there is discarded, not softened.

Not for Google. Google’s guide to optimising for generative AI (May 2026) states explicitly that no AI files, no page fragmentation, no rewriting for AI and no special schema are needed; Google Search ignores llms.txt. augenmerk checks the file but shows it for information only and recommends nothing Google has declared unnecessary. What counts is indexable, useful content and real mentions, and for the other systems the measurement of whether they name your brand.

Via the providers’ APIs with their search tools, and via a measurement service for Google’s answer box. That is not identical to what a logged-in user sees, but it is the only measurement that is reproducible and comparable. We say so openly.

The analysis recognises a question by its wording. A rephrased question starts a new trend from the next run; the old one stays stored. New questions join from the next run; deactivated questions stop the cost but delete nothing. The application checks every question as you type against our question principles.

Nothing. Every question measures all brands in a market at once; another competitor costs no measurement. Only questions cost. The plans limit tracked brands to keep the analysis readable, not because of cost.

Everything augenmerk collects and stores itself sits in a database in Germany. For the measurements, the questions are sent to the AI providers we observe, without account data and without personal data of your customers. Every account sees only its own data.

Create an account, choose a plan, enter brand, website and competitors. We derive the question set from your market, you adjust it, and the first run starts the same week. Payment by card or SEPA direct debit via Stripe; invoices, payment method and cancellation are in your account at any time.