AI Transparency

• The exam agent in the chat is an AI. It asks questions, explains mistakes and accompanies your writing. • The evaluation of your texts (level, points, criteria, feedback) is produced by an AI. • Study plan, practice tasks, model texts, simulations and the estimates in the readiness check (writing, speaking) come from an AI. • The error PDF and the reports in your library are compiled from these AI evaluations. • For teachers: levels and error profiles in the teacher dashboard and the teacher assistant are AI estimates to lighten your preparation, not grades. • Campus articles marked “Created with AI, reviewed by a C1 examiner” are written by an AI and checked by Margarita before publication. You can recognise the AI by the notices where you use it: “You're chatting with an AI.” under the input field, the sentence under every evaluation, and the notice above campus articles. If you ask the agent whether it is human, it will tell you it is an AI.

The didactics, the grading standards and the instructions the AI works by come from Margarita Votteler. The AI carries out what she has laid down. • Margarita Votteler, a C1 examiner since 2009, calibrated the AI evaluation on real exam marking sheets and keeps checking it against texts whose grade is known. • Expert Check and Intensive Week: here Margarita reads your text personally, marks the errors and grades it on the official sheet. That is not an AI. • Campus articles are approved by Margarita before they go live. • If you write to us (hello@germanexam.pro), a human answers.

We use large language models from Google's Gemini family and an openly available model from OpenAI. We reach both through the intermediary OpenRouter. For your personal learning memory we use Zep Cloud. The model that answers can change: if one fails, a backup model steps in. That is why we name model families, not version numbers. Which providers process your texts and how long we store them is explained in the privacy policy. Your texts are not used to train these providers' AI models.

The AI grades against the official criteria of TestDaF, DSH, Goethe C1 and telc C1 Hochschule and shows you the real exam scale plus the CEFR level. It is set up to judge honestly, as on exam day, with no flattery. Its limits: an AI evaluation is a learning aid and can be wrong, especially with very short texts, texts that miss the topic, or unusual formats. It is not an official exam result. Official grades are awarded only by the exam institutions (for example the TestDaF Institute, the Goethe-Institut, telc, and universities for the DSH).

No evaluation here has any legal effect on you. You still have rights: • You can ask at any time for a human to look at your evaluation. Write to hello@germanexam.pro. • You can object to automated processing and use the service for manual learning materials only. • Whether we store your text for quality assurance is your decision, right under the input field, with a one-click objection. • Access, export and deletion of your data: see the privacy policy.

Questions about the AI, the labelling or your rights: hello@germanexam.pro. Responsible: Margarita Votteler, GermanExamPro (see legal notice). This page is dated 4 September 2026. We update it when models or processes change.