Our method
Built by a celiac, because a wrong “safe” can hurt.
YogaNut Life exists for one reason: when you have celiac disease, a mistaken “gluten-free” isn't an inconvenience — it's a physical harm. So we refused to build the easy version, where a chatbot guesses and hopes. Every verdict here is decided by tested code and confirmed by a person.
The founder in the kitchen
The promise: safety is code, not a language model
There's a rule we never break. Any answer that carries real safety consequences — is this product safe to eat? — is produced by deterministic, testable code, not by an AI that predicts plausible-sounding text. AI is wonderful at drafting a friendly explanation. It is exactly the wrong tool for deciding whether a celiac can eat something, because it can be confidently, invisibly wrong.
So we split the work. A rules engine reads the ingredients and decides the verdict. A language model may then draft the explanation around that decision — but it can never change the verdict. And a human reviews it before it's published. Three layers, and the machine that decides safety has no imagination at all. That's the point.
How a verdict is made
- Screen. The ingredient list runs through a published ruleset that looks for wheat, barley, rye, malt, and their derivatives — plus cross-contact risk vectors like non-certified oats.
- Clamp. The result is constrained by the rules: if a gluten grain is present, the verdict cannot be “safe.” This check runs again at the database layer, so even a mistake can't publish an unsafe “safe.”
- Explain. Claude drafts the plain-English rationale around the clamped verdict — prose only.
- Review & source. A celiac reads it, attaches a source (a certification, a manufacturer statement, or the label), and publishes. No source, no publish.
- Monitor. For products we've verified, we re-check the label on a schedule. If a formula quietly changes, the verdict flips to “under review” automatically and members get an alert.
You don't have to take our word for any of it. The exact ruleset is public — read it on the ruleset page.
What we will never do
- Say “100% safe” or “guaranteed.” Formulations change; we tell the truth about uncertainty.
- Let an AI decide a safety verdict.
- Publish a safety claim without a source on file.
- Run fake countdowns or manufactured urgency. When our founding-member spots are gone, the offer simply closes.
Verified information, not medical advice
We're a research and monitoring service, not your doctor. We check products carefully and show our work — but you should always check the current label, and if you feel unwell, contact a medical professional.
Get gluten recall alerts — free.
A celiac reviews every recall before it reaches you. No noise, no fear-mongering. Unsubscribe anytime.