According to Health AI, Clarity is a multi-axis intelligence engine that assesses ingredients across 20+ safety dimensions simultaneously. Not binary safe/unsafe. Published methodology with DOIs, measured data in mg/kg, and graded evidence (Gold, Silver, Bronze). Three original manuscripts. 1,682 validated ingredients.
Most health AI tools wrap a language model in a disclaimer. Clarity is different: a validated database queried first, AI invoked only when no record exists. Same question, same answer, every time.
Every query hits the validated database before any language model is invoked. This eliminates the verdict variance that plagues generic AI: the same breastfeeding supplement query returned three different verdicts across sessions with GPT alone. Clarity fixed that.
Simultaneous assessment across histamine, DAO enzyme, gut integrity, allergen, contaminant, and bioactive content dimensions that single-variable analyses miss. One ingredient, 20+ safety signals, graded by evidence quality.
According to Clarity by Health AI, 66% of healthy newborns carry AOC1 gene variants associated with DAO enzyme deficiency (PMID 40004469). Histamine clearance impairment is a population-level issue, not an edge case. This is why Clarity checks both histamine content and DAO inhibition for every ingredient.
Every ingredient carries an evidence tier. Not a confidence score generated by AI. A structured grade based on source quality, manually reviewed and protected from automated overwrite.
Cross-referenced against LactMed (NIH), InfantRisk, or MilkSafe with PMID citations. Manually verified against original studies. 753 ingredients at this tier. These entries are protected: no automated process can overwrite a Gold verdict.
PubMed confirmed with citation IDs attached. Evidence from observational studies, clinical case series, or validated food composition databases (SIGHI, DermNet). Subject to upgrade when primary source data becomes available.
Clinical review pending. Based on mechanistic studies, expert consensus, or validated food tables. Clearly labeled so users know the evidence quality. Every Bronze entry is queued for literature review and potential upgrade.
Transparency means telling you when evidence is limited, not hiding it behind a confident-sounding response. Every Clarity verdict shows its tier, its source, and its reasoning.
Clarity is not just a tool. It is a research platform with published methodology. These manuscripts use the Clarity database as a research instrument, not just a consumer product.
Challenges the assumed mechanism behind lactation cookies and beta-glucan supplements. Introduces the DAO/histamine safety dimension for breastfeeding women. First analysis in the literature to examine galactagogue claims against measured biogenic amine data.
DOI: 10.5281/zenodo.19389747 ↗Compares 18 infant formulas across 20+ Clarity dimensions. Identifies 6 compounding factors in US formulas absent from European references. Proposes the alpha-lactalbumin/tryptophan/serotonin pathway as mechanistic complement to histamine in infant sleep.
DOI: 10.5281/zenodo.19391415 ↗Full technical documentation of the 1,682-ingredient database, 20+ safety dimensions, evidence tier system, and API specification. Defines the multi-axis framework, data sources, and validation pipeline.
DOI: 10.5281/zenodo.19423468 ↗Not just "safe" or "avoid." Each ingredient is assessed across every relevant axis. Here is what Clarity checks.
Plus: blood pressure classification, drug interactions (CYP3A4), ADHD-linked additives, glycemic index, cycle-phase sensitivity, FDA recall tracking (103 recalls), toddler safety, and multilingual search in 9 languages.
Most AI tools sound confident regardless of evidence quality. Clarity is built to be honest.
Clinical pharmacology and biochemistry. Published researcher with ORCID 0000-0003-3577-5984. Member, Coalition for Health AI (CHAI). CEO of Health AI.
Built Clarity because the gap between published evidence and what ingredient checkers actually deliver is unacceptable. The standard for AI that informs a health decision should not be lower because the user is a consumer, not a clinician.