01
When AI is involved
Summaries, quizzes, flashcards, chat, Lesson Mode, images and suggestions may be created or transformed using AI systems. The interface and exports identify them as AI-assisted content.
AI transparency & responsible use
Know when AI helps — and where you stay in control.
Memoniq is an AI-assisted self-study tool. Generated content can be useful and still be wrong: compare important claims with your original sources before relying on them.
02
Human control
Memoniq supports self-study. Recent practice and immediate retries do not certify retention or exam readiness. Check sources retrieves related notebook excerpts, not confirmed citations or automatic fact checks. Open the original document and check its context. Memoniq does not issue official grades or decide access to education.
03
Data minimisation
Only content needed for the requested feature is sent to providers. Avoid health or biometric data, identity documents, children's data or unauthorised third-party information.
04
Reports and rights
You can export or delete your data in settings, revoke shared links, and report incorrect or risky output. Privacy requests use the contact listed in the privacy notice.
Intended purpose
Learning support, not automated educational decisions.
Practice scores, oral-exam feedback and mastery indicators are estimates for personal study. They must not be used as authoritative grades, admission, placement, disciplinary or proctoring decisions.
Where AI creates or transforms content, Memoniq adds visible disclosure and provenance metadata to supported exports. Legacy files may not contain the same markers.
You are interacting with AI systems
These Memoniq features use AI. When you chat with the tutor or answer the oral examiner, you are talking to an AI system, not a person.
- Summaries, “Must know” sections and text optimisations
- Quizzes, mock exams and flashcards
- Notebook tutor chat and explanations in lessons
- AI oral examiner (questions, answer evaluation, feedback)
- Podcasts with AI scripts and synthetic voices
- Lessons and exported videos with synthetic narration
- Diagrams generated from the summary, image selection and checks
- Transcription of audio, video and oral answers; reading images and scanned PDFs
- Linking summary sentences to sources and claim verification
Models and providers per feature
Managed defaults as of September 28, 2026. They may change for quality, cost, security or availability; with your own API key or a local model, the provider you configure is used instead.
| Feature | Model / provider |
|---|---|
| Text: summaries, quizzes, flashcards, chat, oral examiner, scripts | DeepSeek V4.1 Flash (DeepSeek) via OpenRouter; some coverage checks use GPT-6 Luna (OpenAI). Fallback: Google Gemini, Groq, OpenAI, MiniMax |
| Claim verification | Jev (verification model) via OpenRouter, escalating to DeepSeek; bibliographic search on PubMed |
| Reading images and scanned PDFs | Qwen3-VL (Alibaba Qwen) via OpenRouter, Google Gemini; Tesseract OCR running in your browser |
| Transcription | Whisper Large v3 Turbo (OpenAI) via OpenRouter; fallback Groq Whisper and Google Gemini |
| Synthetic voice (podcasts, lessons, videos) | Fish Audio S2.1 Pro (free version) via OpenRouter; fallback Kokoro 82M; then Microsoft Edge voice in the browser. Optional local voices (Kokoro, Piper) |
| Linking to sources | text-embedding-3-small (OpenAI) via OpenRouter |
| Diagrams | Diagram code written by the text model, drawn by Kroki (no AI in the drawing) |
| Generated images | FLUX.2 klein (Black Forest Labs) and Gemini 2.5 Flash Image (Google) via OpenRouter, currently internal preview only |
| Mind map | No AI: built deterministically from the summary |
Synthetic audio and video
All podcast voices and all narration in lessons and exported videos are AI-generated synthetic voices, not recordings of real people. The interface labels them, exported videos carry a visible AI notice, and exported videos (MP4/WebM), audio downloads (MP3/WAV), generated images and DOCX, PPTX and Markdown exports include machine-readable AI provenance metadata; Anki exports carry an “ai-generated” tag. PDF exports carry the visible notice (a PDF printed from the browser does not keep the metadata). If you publish exported content, you must keep the notice and disclose that it is AI-generated.
Scores, voice and what we do not do
Quiz and oral practice scores are practice estimates based on the content of your answer. Memoniq does no emotion recognition and no biometric analysis of voice or face. Your spoken answers are transcribed to text and the audio is not stored: we do not keep voiceprints. Only per-topic scores and short notes are saved.
Report an AI error
Use the “Report” link next to the AI notices in the app, or write to infomemoniq@gmail.com with the notebook, the feature and what is wrong. Reports are reviewed by a person and used to fix prompts, models or sources.
EU AI Act classification
Memoniq is a general-purpose educational tool for voluntary self-study. It is not intended or offered for admission, access to education, evaluation of learning outcomes that determine grades or credentials, placement or exam proctoring. On this basis we consider it not high-risk under Annex III(3) of Regulation (EU) 2024/1689, and we apply the Article 50 transparency obligations (disclosure of AI interaction and marking of synthetic content). No practice prohibited by Article 5, such as emotion recognition in education, is used.
Official framework
These product measures support transparency and data protection by design; they are not a legal certification. Organisational records, vendor agreements, DPIA where required, training and legal review remain necessary.