Editorial Policy
Last updated: February 2026
Our approach
Every score, recommendation, and AI response here is built to inform you, never to replace professional medical advice. You deserve to know exactly how your food is evaluated, and why. That is what this page is for.
How content is created
- PQ Score methodology. Our team built the Pure Quality scoring algorithm and keeps it updated, informed by peer-reviewed research on food processing, ingredient safety, and dietary quality. Every penalty category, seed oils, added sugars, artificial additives, ultra-processing markers, is documented with the reasoning behind it.
- AI-generated coaching. Coach's responses come from large language models (Gemini and OpenAI), run through prompts built to enforce evidence-based nutrition principles. Guardrails at the system level block medical claims, push back on extreme restriction, and flag when you should talk to a professional instead.
- For You personalization. Health-context scoring, conditions, allergens, dietary preferences, is built on documented research and conservative flagging rules. When we are not sure, we flag it rather than stay quiet.
- Nutritional data. Food composition data comes from USDA FoodData Central, FatSecret, and Open Food Facts. Our sourcing policy has the full breakdown.
Transparency standards
- Score breakdowns. Every PQ Score shows exactly which ingredient categories caused deductions. No black boxes.
- Confidence indicators. When data is incomplete, a food search without a full ingredient list, we say so with a lower confidence indicator instead of guessing.
- No paid placements. Brand partnerships, advertising, and sponsorships never touch a food score or a recommendation. We do not take payment to change either one.
- Corrections. Think a score or recommendation is wrong? Report it in the app. We look into it and update our scoring rules when the evidence supports it.
How we cite research
When we refer to research, we link to it. Every reference carries a permanent link to the published paper, and we say what the study actually measured, including what it does not show. Where we cannot point to a specific source, we say so instead of implying one exists. The references below are the published work behind the ingredient-quality reasoning in the PQ Score.
Sources behind PQ Score ingredient reasoning
- Chassaing B, Koren O, Goodrich JK, Poole AC, Srinivasan S, Ley RE, et al. (2015). Dietary emulsifiers impact the mouse gut microbiota promoting colitis and metabolic syndrome. Nature, 519(7541), 92 to 96. doi.org/10.1038/nature14232 PubMed 25731162
The specific 2015 Nature finding the scanner refers to when it flags emulsifiers: in mice, dietary emulsifiers altered the gut microbiota and promoted colitis and metabolic syndrome. What it does not show: Conducted in mice. It is the reason we flag emulsifiers as worth knowing about, not evidence that they will do this in any person.
- Witkowski M, Nemet I, Alamri H, Wilcox J, Gupta N, Nimer N, et al. (2023). The artificial sweetener erythritol and cardiovascular event risk. Nature Medicine, 29(3), 710 to 718. doi.org/10.1038/s41591-023-02223-9 PubMed 36849732
Higher circulating erythritol levels were ASSOCIATED with a higher rate of major adverse cardiovascular events. This is the study behind the erythritol caution in scan results. What it does not show: Association, not causation, and it measured blood levels rather than dietary intake. It does not establish that eating erythritol causes cardiovascular events.
- Bikman BT, Summers SA (2011). Ceramides as modulators of cellular and whole-body metabolism. The Journal of Clinical Investigation, 121(11), 4222 to 4230. doi.org/10.1172/JCI57144 PubMed 22045572
A review setting out ceramides as modulators of insulin signaling and whole-body metabolism. This is the specific published work behind the "insulin is central to metabolic health" framing used on the site. What it does not show: A mechanistic review, not a clinical trial. It does not establish that any particular diet produces any particular outcome in any particular person.
- Holland WL, Bikman BT, Wang LP, Yuguang G, Sargent KM, Bulchand S, et al. (2011). Lipid-induced insulin resistance mediated by the proinflammatory receptor TLR4 requires saturated fatty acid-induced ceramide biosynthesis in mice. The Journal of Clinical Investigation, 121(5), 1858 to 1870. doi.org/10.1172/JCI43378 PubMed 21490391
The saturated fat to TLR4 to ceramide to insulin resistance pathway that the app’s ingredient-quality reasoning refers to when it treats fat SOURCE as meaningful rather than treating all fat as interchangeable. What it does not show: Conducted in mice. It is mechanistic evidence for a pathway, not evidence of a human dietary outcome.
What CleanKeto is not
CleanKeto is a nutritional information tool, not a medical device. We do not diagnose, treat, or prescribe. The PQ Score and the For You score are CleanKeto's own proprietary editorial opinion of ingredient quality and personal fit. They are informational only, they are not a safety rating, and they are not a substitute for professional dietary or clinical guidance. Always consult qualified healthcare professionals for medical decisions.
Review process
Our scoring methodology undergoes periodic review. Major changes to penalty weights, ingredient classifications, or AI coaching guidelines are tested against adversarial evaluation suites before deployment. We maintain version history for all scoring changes.