AI Transparency

Rizzky is built on large language models from Google (Gemini API). We combine these models with our own instructions, settings and app design and offer the result under the Rizzky name. That makes Tobias Scharl (trading as Rizzky) the provider of the Rizzky AI system under the AI Act. Google remains responsible for the underlying general-purpose models.

Models currently in use: Gemini 3.5 Flash-Lite (reading screenshots) and Gemini 3.7 Flash (coaching, replies and AI chats). When you upload a screenshot, two steps happen:

1. Reading the screenshot. A vision model turns the image into structured text: who said what, in which order, and the visible text of a profile (bio, prompts, captions). This step only extracts text. It does not judge the people in the image.
2. Writing the suggestions. A language model uses that text, your messaging-style settings and, if you set one up, the notes from an AI chat for that person. It returns an interest score (for conversations), a short coaching read and several reply suggestions in different tones, or opening lines for a profile.

AI chats work the same way: you type questions about a situation and the model answers, using what you told it when you set up the chat and the latest reading for that person.

• It does not send, schedule or post anything. Nothing leaves Rizzky unless you copy it and send it yourself.
• It does not analyze faces, bodies, voices or any other biometric data. Photos in a screenshot are not used to judge looks, emotions, age, ethnicity or sexual orientation.
• It does not use emotion recognition in the sense of the AI Act. The interest score comes from written text only.
• It does not make decisions about you with legal or similar effects, and it doesn't decide who you match or talk with.
• It does not build profiles of the people in your screenshots or connect them across different users.

The interest score is a rough estimate of how engaged the other person seems in the conversation you uploaded. The model looks at signals in the text, such as how long and how detailed their replies are, whether they ask questions back, whether they pick up on what you said, who starts topics, and whether plans come up. It weighs these into a score and explains the main reasons in the coaching read.

The score only sees the screenshot you share. It doesn't know about calls, meetings, earlier chats you didn't upload, or anything going on in the other person's life. Treat it as a second opinion, not a verdict. The quizzes on our website work differently: they are simple scoring rules, not AI.

AI language models are good at tone and phrasing, and bad at certainty. In practice this means:

• Suggestions can misread the situation, miss sarcasm or inside jokes, or sound off for you or the other person.
• Text extraction can make mistakes with unusual layouts, small fonts, emojis or several languages in one chat.
• The models learned from large amounts of text and can reflect stereotypes. We tell them to avoid this, but we can't rule it out.
• Rizzky gives dating and texting suggestions. It is not therapy, medical, legal or safety advice.

If you are worried about your safety or someone else's, please contact local emergency services or a support service near you instead of relying on an app.

Every suggestion is a draft. You can regenerate it, edit it, ignore it or write your own reply. You are responsible for what you choose to send, and you should only send messages you would be comfortable sending without Rizzky. Please don't use Rizzky to pressure, deceive or manipulate anyone. Our Terms of Service prohibit that.

To create a reading, the extracted text and your settings are sent to Google (Gemini API). We do not use your screenshots, your conversations or the people in them to train AI models, and our AI provider is not allowed to use them for training either. Screenshot images are never saved (they are only held in working memory while the text is extracted).

Details about what we store, for how long, where it is processed and your rights are in our Privacy Policy.

The free tools on this website (the quizzes, the rizz test and the line, bio, Hinge prompt and date idea generators) do not use AI. They pick from lines and ideas our team wrote and combine them with simple rules, right in your browser. Nothing you type into them is sent to us.

• Rizzky is subject to the EU AI Act (Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744). Its transparency rules for AI systems apply from 2 August 2026.
• Rizzky is not a high-risk AI system under the AI Act. It does not fall into any of the areas listed in Annex III and does not use biometric data.
• We train our team on how the AI works and where it fails, so they can handle questions and reports properly (Art. 4 AI Act).
• The market surveillance authority for AI in Germany is the Federal Network Agency (Bundesnetzagentur).

If a suggestion is offensive, harmful or plainly wrong, please tell us. Use the "Report AI content" link at the bottom of a result or AI chat, or email rizzkyapp@gmail.com and, if possible, include what happened. Reports are read by a person and help us improve the instructions we give the models. For anything else, see our Imprint.