AI Use & Limitations

We use AI to help you make sense of your own health data. Specifically, it:

• reads and structures uploaded documents (labs, imaging, genetic, microbiome, ECG);
• explains what a marker is and what a result may mean, in plain language;
• answers your questions in the assistant, drawing on your files and our knowledge base;
• surfaces informational signals, such as values that sit outside a typical range.

AI-generated content is produced by statistical models and can be inaccurate, incomplete, outdated, internally inconsistent, or simply fabricated (“hallucinated”). It may misread a document, miss context about your specific situation, or reflect information that is no longer current. All AI output is provided for information only and “as is,” with no guarantee of accuracy.

Do not use LifeFrom’s AI to diagnose yourself, to decide whether to seek care, or to start, stop, or change any medication or treatment. It is not a clinician and does not provide medical advice. Always confirm anything that matters with a qualified healthcare professional. See our Medical Disclaimer.

Many explanations draw on an internal, reviewed knowledge base of biomarker information, with sources cited where shown. This improves consistency and traceability, but it is still general educational content, not advice tailored to you, and it may not reflect the very latest evidence or your individual circumstances.

AI is optional. LifeFrom works without it. If you turn LifeFrom AI off in Settings, all AI processing stops and nothing more is sent to AI providers.

When it is on, and only after you agree in the app, these providers process what is needed to answer you:

• OpenRouter: the gateway every AI request goes through. We ask it on every call to route only to providers that don't train on the data and that offer zero data retention where available;
• OpenAI models, which may be served through Microsoft Azure via OpenRouter, for answers and reading results;
• Google Gemini models, for answers and reading photos and documents;
• Mistral, for text recognition (OCR) on documents you upload;
• Deepgram, for turning your voice into text when you dictate on the website;
• an image model, which receives only a dish name, to draw food illustrations.

Our requests are set not to train on your data, and to use zero-retention endpoints where available; some providers may keep a request for up to 30 days for abuse monitoring. When LifeFrom looks up research, it sends search terms only, never your personal data, to PubMed, Europe PMC, Semantic Scholar and ChEMBL.

We do not use your identifiable health data to train AI models. You can delete your data at any time in Settings, or email privacy@lifefrom.ai for a copy. Full details are in our Privacy Policy.

We don’t use AI to make decisions that produce legal or similarly significant effects about you without human involvement. You remain in control of your account and your data, and you decide what to do with any information the AI provides.

If an AI response looks wrong, confusing, or harmful, please tell us at support@lifefrom.ai. Feedback like this directly helps us improve the safety and quality of the Services.

Questions about how we use AI? Email privacy@lifefrom.ai.