The Vera Purity Index.

A composite score from 0 to 100 that rates bottled water quality across seven independently weighted dimensions. Designed to go beyond regulatory compliance and reflect what the peer-reviewed literature suggests is optimal for human health.

How it starts

Every product starts at 50, representing a baseline water with no distinguishing qualities in either direction. Points are added or subtracted based on seven categories of evidence. The final score is clamped between 0 and 100. This lower starting point makes high scores harder to achieve and the scale more honest overall.

Score tiers

80 - 100Excellent
60 - 79Good
40 - 59Fair
20 - 39Poor
0 - 19Very Poor

1. Contaminant Safety

-40 to +15

Evaluates the concentration of known chemical contaminants against two separate benchmarks: the U.S. EPA Maximum Contaminant Levels (MCLs), which are the legal limits, and the Environmental Working Group (EWG) Health Guidelines, which are set at stricter levels based on one-in-a-million cancer risk or no-observed-adverse-effect levels.

Why two benchmarks? EPA MCLs are set with economic and technological feasibility in mind, not purely on health grounds. EWG guidelines are derived solely from health-protective science, and for many contaminants are substantially lower than the legal limits.

Scoring logic

  • Exceeding an EPA MCL: severe penalty scaled by log-ratio and toxicity weight
  • Exceeding an EWG guideline: moderate penalty scaled by ratio and toxicity
  • Approaching an EWG guideline (50-100% of limit): minor penalty
  • Below 50% of EWG guideline: no penalty
  • All below 10% of EWG: +15 | Below 25%: +10 | Below 50%: +5

Penalty formulas & log scaling

EPA MCL breach

penalty = weight x (5 + log10(measured / limit) x 10)

EWG guideline breach

penalty = weight x (2 + log10(measured / guideline) x 5)

Toxicity weight:

The 40-point cap (dashed line) prevents any single contaminant from dominating the total score. Log scaling compresses extreme values: 10x the limit only adds ~30 pts more than being exactly at the limit.

Worked examples

Arsenic at EPA MCL

10 ppb (exactly at limit)

ratio = 1.0 | log10(1.0) = 0.000

penalty = 3 x (5 + 0.000 x 10) = 15 pts

Arsenic at 2x EPA MCL

20 ppb

ratio = 2.0 | log10(2.0) = 0.301

penalty = 3 x (5 + 0.301 x 10) = 24 pts

Arsenic at 10x EPA MCL

100 ppb

ratio = 10.0 | log10(10.0) = 1.000

penalty = 3 x (5 + 1.000 x 10) = 40 pts

Lead above EWG

3 ppb (EWG guideline = 1 ppb)

ratio = 3.0 | log10(3.0) = 0.477

penalty = 3 x (2 + 0.477 x 5) = 13.2 pts

Why logarithmic, not linear? A linear penalty would make arsenic at 10x the limit cost 10x more than being exactly at the limit. The log scale compresses this: going from 1x to 10x the limit only adds ~30 points (15 to 45, capped at 40). This prevents a single extreme outlier from collapsing the entire score, while still penalising dangerous exceedances heavily.


Toxicity weighting

ContaminantWeightBasis
Lead, Arsenic3.0IARC Group 1 carcinogens; no safe threshold
Chromium-6, Uranium2.5Known/probable human carcinogens
PFAS compounds2.5Persistent endocrine disruptors; EPA 2024 MCL at 4 ppt
Mercury, Bromate, THMs, HAAs2.0IARC Group 2A/2B; cardiovascular/renal toxicity
Nitrate, Nitrite, Fluoride, Cadmium1.5Developmental toxicants; dose-dependent risk
Barium, Copper, Chlorine0.5-1.0Lower systemic toxicity at typical concentrations
Key references

2. Transparency & Testing

-20 to +15

A brand's willingness to publish independent test data is itself a signal of quality and accountability. This category rewards comprehensive testing and published results.

Required test categories (9 total)

01Heavy metals
02Disinfection byproducts
03Volatile organic compounds
04Semi-volatile organic compounds
05Microbiological indicators
06Fluoride
07Radiological contaminants
08Nitrate / nitrite
09Disinfectant residuals

Scoring logic

ConditionScore
No test data at all-10
Some data (lab reports or minerals, no formal categories)-5
Partial coverage (<50% of required categories)-5
Moderate coverage (50-79%)-3
Strong coverage (80-99%)+3
Full coverage (all 9 categories)+7
Published lab reports (1 report)+3
Published lab reports (2+ reports)+5
Recent reports (<2 years old)+3
Third-party laboratory verification+5
NSF International certification+4
Key references

3. Source & Treatment

-10 to +10

The origin of water and how it is processed before bottling meaningfully affects its purity, mineral integrity, and exposure to industrial contamination.

Source adjustments

Source TypeScoreRationale
Spring / Aquifer+5Natural filtration through geological substrate
Deep well / Glacier / Iceberg+3Remote, low-exposure sources with natural mineral content
Rain / Atmospheric-3Susceptible to atmospheric pollutant deposition
Municipal (tap-sourced)-5Pre-treated municipal supply; chemical disinfectants may persist
Unknown / undisclosed-10Lack of provenance is itself a transparency concern

Treatment adjustments

TreatmentScoreRationale
UV disinfection+3Pathogen inactivation without chemical residues
Ozonation+2Effective disinfection; breaks down to oxygen
Activated carbon filtration+2Reduces VOCs, chlorine, and taste compounds
Reverse osmosis0Effective purification but removes beneficial minerals
Chlorination-2May leave residual trihalomethanes and chlorine
Chloramine-3Harder to remove; associated with disinfection byproduct formation
Key references

4. Packaging & Materials

-10 to +5

Container material is a major but often overlooked determinant of bottled water quality. Plastics can leach monomers, plasticisers, and other chemicals into water, particularly under heat or UV exposure.

MaterialScoreRationale
Glass+5Chemically inert; zero leaching
Stainless steel+5Inert, durable; no documented leaching
PLA bioplastic-1Compostable; low leaching risk but limited data
Cardboard / Tetra Pak-2Low leaching; some adhesive layer concerns
Aluminum-3Typically lined with epoxy/BPA-based coating
HDPE (#2)-3Lower leaching than PET; used for large-format jugs
Polypropylene (#5)-4Moderate leaching at elevated temperatures
Polyethylene-5Variable additive packages; moderate risk
PET (#1)-7Leaches antimony, acetaldehyde; risk increases with heat/reuse
Polystyrene (#6)-8Leaches styrene monomer; IARC Group 2B carcinogen
PVC (#3)-10Leaches vinyl chloride, phthalates, lead stabilisers

Certified microplastic-free: penalty halved

BPA-free certification on aluminum: +2

Key references

5. PFAS & Emerging Contaminants

-10 to +5

PFAS and microplastics represent two of the most significant emerging concerns in drinking water science. Unlike legacy regulated contaminants, these were not routinely tested for until recently, anda brand's decision to test and disclose results is itself meaningful.

PFAS scoring

ResultScoreRationale
Not tested0Neutral; no data either way
Non-detect+5Demonstrates active quality management
Detected, below EWG guideline-3Concerning presence of persistent chemicals
Detected, above EWG guideline-10Significant health risk; EPA 2024 MCL = 4 ppt

Microplastics scoring

ResultScore
Not tested0
Non-detect+2
Detected, <=10 particles/L0
Detected, >10 particles/L-5
Why PFAS matters

PFAS are synthetic fluorinated compounds persistent in the environment and the human body, linked to thyroid disruption, immune suppression, reproductive toxicity, and several cancers. In 2024, the EPA established the first-ever federal MCL at 4 parts per trillion (ppt) for PFOA and PFOS.

Why microplastics matter

A 2018 Orb Media analysis found microplastic contamination in 93% of tested bottled water brands globally. A 2021 study estimated humans ingest 39,000-52,000 microplastic particles annually. Toxicological implications include inflammation, endocrine disruption, and carriage of adsorbed chemical pollutants.

Key references

6. Mineral Profile

-10 to +20

The only category that can award more points than it deducts, reflecting the growing evidence that mineral-rich water confers measurable health benefits beyond simple hydration. Each mineral is scored individually against evidence-based ranges, followed by a diversity bonus.


Calcium

Associated with reduced cardiovascular mortality and improved bone mineral density. WHO recommends a provisional minimum of 20 mg/L, with strongest cardiovascular benefit at >50 mg/L.

Calcium (mg/L)ScoreLabel
>= 50+4High — supports bone & cardiovascular health
20 - 49+3Good
10 - 19+1Adequate
5 - 90Neutral
< 5-1Very low

Magnesium

Essential cofactor in over 300 enzymatic reactions. Low dietary magnesium is among the most consistent risk factors for cardiovascular disease. Water-borne magnesium is highly bioavailable relative to food sources.

Magnesium (mg/L)ScoreLabel
>= 30+4High — strong cardiovascular & muscle benefit
10 - 29+3Good — supports cardiovascular health
5 - 9+1Adequate
< 5-1Very low — associated with higher cardiovascular risk
> 125-2Very high — laxative effect risk

Silica

The PAQUID cohort study (7,598 French adults, 15 years) found silica intake >= 10 mg/L was significantly associated with reduced dementia incidence. Silicon-rich water (>= 30 mg/L) measurably reduced urinary aluminium excretion, suggesting protective role against aluminium neurotoxicity.

Silica (mg/L)Score
>= 30+4
10 - 29+3
5 - 9+1

Bicarbonate

Primary physiological pH buffer. Waters with >= 150 mg/L have been associated with improved acid-base balance and reduced cardiovascular risk markers in controlled trials.

Bicarbonate (mg/L)Score
>= 150+2
50 - 149+1

Sodium

While water is a minor sodium source for most people, those on sodium-restricted diets should monitor water sodium. WHO's palatability guideline of 200 mg/L is aesthetic only.

Sodium (mg/L)ScoreLabel
<= 20+2Naturally low
21 - 50+1
> 100-1Elevated
> 200-3Exceeds WHO palatability guideline

Fluoride

Exhibits a classic hormetic dose-response: low concentrations protect dental enamel; above certain thresholds it causes fluorosis. The 2015 U.S. Public Health Service optimal level is 0.7 mg/L.

Fluoride (mg/L)ScoreLabel
<= 0.3+1Very low; negligible fluorosis risk
0.3 - 1.00Near optimal
1.0 - 1.5-1Above CDC optimal
1.5 - 2.0-3Exceeds WHO health guideline
2.0 - 4.0-4Dental fluorosis risk
> 4.0-6Exceeds EPA MCL; skeletal fluorosis risk

pH

The health significance of pH is frequently misunderstood. WHO's 6.5-8.5 guideline is aesthetic, not health-based. Commercially marketed "alkaline water" (pH 8-10, artificially raised) has been found in RCTs to produce no clinically meaningful health effect.

pHScoreNote
< 5.5-5Significant dental erosion risk
5.5 - 6.5-3Mild erosion risk
6.5 - 9.50Optimal range; no evidence of harm or benefit
> 9.5-3Extreme alkalinity; case reports of harm
> 8.5, artificially alkalised-2No evidence of benefit

TDS (Total Dissolved Solids)

A summary measure of all dissolved minerals, used as a cross-check signal rather than a primary determinant.

TDS (ppm)ScoreNote
< 20-4Essentially distilled; mineral-free
20 - 50-2Very low; heavily filtered or RO
50 - 1000
100 - 600+1WHO "excellent" to "good" palatability range
600 - 1200-1WHO "fair" palatability
> 1200-2WHO "unacceptable"; high dissolved solids

Mineral Diversity Bonus

The combination of calcium, magnesium, silica, and bicarbonate likely produces synergistic cardiovascular and metabolic benefits beyond any single mineral in isolation.

Minerals in beneficial rangeBonus
4+ minerals+5
3 minerals+3
2 minerals+1

Remineralisation caveat: Products artificially remineralised after reverse osmosis receive 50% of their calculated mineral score. While resulting concentrations may be identical to a natural spring, the bioavailability of naturally occurring mineral complexes is superior to post-processing additions.

Key references

7. Brand Track Record

-15 to +5

Track Record measures one thing: does this brand have a clean legal and safety history? It is deliberately kept separate from Transparency & Testing, which rewards published lab data. Previously, this category gave bonus points for multiple lab reports — but that double-counted with Transparency. Now the two categories measure completely different signals:

Transparency

Do you publish your testing data?

Track Record

Do you have a clean legal/safety history?

FactorAdjustment
Product recall (1 event)-5
Multiple recalls-10
California Prop 65 warning-5
Verified legal action (contamination / fraud)-5
Clean record (zero recalls, lawsuits, or Prop 65 warnings)+5

Recalls sourced from U.S. FDA Enforcement Reports. Prop 65 listings from California OEHHA. Legal actions included only with verifiable source URL.

How Scoring Works

AI Research & Human Verification

The Automated Pipeline

When you scan a product that isn't yet in our database, Vera's research pipeline activates automatically. Within 30–60 seconds, the system runs the following steps in parallel:

1

AI Knowledge RecallClaude (Anthropic)

Queried for everything it knows about the product: source location, mineral composition, pH, TDS, treatment methods, packaging, testing history, and brand track record. For well-known brands, Claude’s training data often contains accurate mineral profiles sourced from brand documentation, regulatory filings, and scientific literature.

2

Web SearchTavily

Eight independent searches run simultaneously, targeting official lab reports, water quality certificates, mineral analyses, regulatory databases (EWG, NSF), consumer watchdog coverage, and legal records. Discovered PDF reports and quality pages are fetched directly.

3

Structured ExtractionGemini (Google)

All gathered content is passed to Gemini, which extracts a standardised data structure covering over 40 fields: mineral concentrations, contaminant levels, source coordinates, treatment methods, packaging type, PFAS testing status, microplastic data, and track record events.

4

ScoringVera Purity Index Engine

The extracted data is passed through the scoring engine, which calculates the score across all seven categories using the methodology described above.

5

StorageInstant availability

The product profile, mineral data, contaminants, lab report links, and score are stored and immediately available to all users.

This entire process runs without human involvement.


The Confidence Score

Because the automated pipeline relies on AI and publicly available web data, not every product is equally well-documented. A niche regional brand with no English-language web presence will naturally have less verifiable data than a global brand with published annual quality reports and NSF certification.

Every AI-researched product carries a confidence score from 0 to 100, displayed on the report. This score is self-assessed by the AI extraction model based on the quality and diversity of sources it was able to draw from:

90 – 100%

Multiple corroborating official sources — PDFs, regulatory databases, NSF/EWG records — with specific numeric values cross-checked across sources.

70 – 89%

Good web source coverage combined with AI knowledge; most key fields populated with real, verifiable values.

50 – 69%

Primarily AI training knowledge with limited or no independent web verification; figures are plausible but unconfirmed.

30 – 49%

Limited data available; most fields estimated or inferred; treat results with caution.

0 – 29%

Almost no verifiable data found; product is largely unknown to public databases.

A low confidence score does not mean the water is unsafe — it means we simply don't have enough data to evaluate it thoroughly. In these cases, the absence of transparency is itself reflected in the Transparency & Testing category score.


The Verified Badge

The Verified badge is awarded only after a member of the Vera team has manually reviewed the product report against primary sources. Verification involves:

  • Cross-checking mineral and contaminant values against the brand’s official published documents or third-party lab reports
  • Confirming the source, treatment methods, and packaging type
  • Reviewing any flagged concerns (recalls, lawsuits, Prop 65 listings) against primary records
  • Adjusting the score if the automated extraction produced inaccuracies

An AI-researched report without the Verified badge should be treated as a best-effort estimate, not a certified finding. The confidence score gives you a signal of how much to trust it. A report with 85% confidence and coherent mineral data is likely reliable. A report with 35% confidence should be read with healthy scepticism until verified.

We are progressively verifying products in the database, prioritising those with the most scans and lowest confidence scores. If you believe a report contains an error, you can flag it directly from the product page and our team will review it.


A Note on Limitations

No automated system is perfect. AI models can hallucinate specific values — particularly for obscure products with little web presence. Web searches may surface outdated lab reports. Brands occasionally update their source, treatment process, or formulation without public announcement.

We build these limitations directly into the scoring system: brands that do not publish test data receive transparency penalties, and products without verified human review carry the confidence score as an explicit uncertainty signal. Our goal is not to create false precision, but to give you the most honest and useful picture of water quality that current data allows.

Limitations & Disclosures

The VPI is a scoring model, not a regulatory certification. It synthesises available published evidence and disclosed product data.

Where brands do not publish contaminant test data, scores reflect that absence of transparency rather than confirmed quality.

Mineral and contaminant values reflect conditions at the time of last available testing. Water quality can vary between batches, seasons, and production facilities.

The VPI does not test products independently. Where brands are found to have published inaccurate data, scores will be updated.

Vera Purity Index methodology current as of 2026. Last reviewed February 2026.