BRENNEN

Industries / Food, farming and forestry / Food and beverage processing / Bottled water

Quality testing

Quality testing is the checking of bottled water, at the source, on the filling line and after it has sat on a shelf, against legal limits for germs, chemicals and anything that seeps out of the bottle.

7

Milestones achieved, from 1995 to 2025.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1995

    US defines bottled water

    A US Food and Drug Administration (FDA) rule of 13 November 1995 gave legal meanings to names such as spring water, artesian water, purified water and mineral water, which must hold at least 250 parts per million of dissolved solids, and set allowable levels of contaminants in bottled water. As amended since, it caps arsenic at 0.010 mg/L and lead at 0.005 mg/L. FDA's manufacturing rules require a plant to test its product for coliform bacteria at least once a week and its source water for chemical contaminants at least once a year.

    law.cornell.edulaw.cornell.edulaw.cornell.edu

    Credit

    US Food and Drug Administration

  2. 2006

    Antimony leaches from PET

    Antimony is a toxic metal; PET is the clear plastic of most water bottles. Using clean-room methods built for polar ice, a University of Heidelberg team found 2.2 ng/L of antimony in pristine Ontario groundwater but an average of 156 ng/L in 12 Canadian brands bottled in PET. One German water held 3.8 ng/L before bottling, 359 ng/L in PET bottles from a shop, and 626 ng/L after three more months on a shelf. All these levels are far below FDA's limit of 0.006 mg/L (6,000 ng/L).

    Lessonthe container is part of the water's chemistry, so a test at the source does not describe what the drinker gets.

    europepmc.org

    Credit

    W. Shotyk, M. Krachler and B. Chen, University of Heidelberg

  3. 2009

    Auditors compare bottle and tap

    The US Government Accountability Office (GAO) reported that FDA's safety and consumer protections for bottled water "are often less stringent than comparable EPA protections for tap water". On its recommendation, FDA set a limit of 0.006 mg/L for DEHP, a chemical that softens plastics, in a final rule of 19 October 2011. GAO's second recommendation, to give buyers more information, is closed as not implemented; FDA cited limited resources.

    Lessonan outside audit got a new limit set, but not better information for buyers.

    gao.gov

    Credit

    US Government Accountability Office; US Food and Drug Administration

  4. 2018

    Microplastics in 93% of bottles

    Researchers at the State University of New York at Fredonia tested 259 bottles of 11 brands, among them Aqua, Bisleri, Wahaha, E-Pura and Nestlé Pure Life, bought in nine countries. 93% held some microplastic, plastic pieces smaller than 5 mm. They counted an average of 10.4 particles per liter larger than 100 µm, and 325 per liter when smaller pieces, down to 6.5 µm, were included; the smaller ones were stained with a dye but not chemically confirmed. Polypropylene, the usual plastic of bottle caps, made up 54% of the confirmed particles.

    Lessoncounts depend on the smallest size a method can see, so the World Health Organization (WHO) said in August 2019 that standard measuring methods were needed.

    frontiersin.orgwho.int

    Credit

    Sherri Mason, Victoria Welch and Joseph Neratko

  5. 2024

    Nanoplastics counted one by one

    A Columbia University team built a stimulated Raman scattering microscope, which identifies a plastic by how its molecules vibrate under laser light, that can pick out single plastic particles below 100 nm. It estimated about 240,000 plastic particles per liter of bottled water, about 90% of them nanoplastics (smaller than 1 µm), which the authors say is orders of magnitude more than earlier counts. In February 2026 an Ohio State University team, using electron microscopy with infrared spectroscopy, reported about three times as many nanoplastic particles in some bottled brands as in treated tap water from four plants near Lake Erie.

    europepmc.orgnews.osu.edu

    Credit

    N. Qian, W. Min and colleagues (Columbia); Megan Jamison Hart and John Lenhart (Ohio State)

  6. 2024

    India moves to yearly inspections

    India's food rules had required packaged drinking water and mineral water to carry the certification mark of the Bureau of Indian Standards (BIS), which had granted over 6,200 licences for packaged drinking water by September 2022. An amendment notified on 17 October 2024 removed that requirement. On 29 November 2024 the food regulator, FSSAI, put both products in its high-risk category instead: makers are inspected before they get a licence and at least once a year, and makers with central licences must pass a yearly audit by an FSSAI-recognised third party. ThePrint reports that nearly 80% of India's industry is unorganised.

    fssai.gov.inhealth.medicaldialogues.inservices.bis.gov.intheprint.in

    Credit

    Food Safety and Standards Authority of India

  7. 2025

    FDA tests bottles for PFAS

    PFAS (per- and polyfluoroalkyl substances) are man-made fluorinated chemicals that persist in the environment. FDA tested 197 bottled waters bought at US shops in 2023 and 2024 for 18 PFAS; 10 had detectable levels, and none would have exceeded EPA's limits for tap water.

    fda.gov

    Credit

    US Food and Drug Administration, April 2025

Watch for

  1. PFAS limits for bottled water

    EPA set PFAS limits for tap water in April 2024 and proposed amending them in May 2026. When EPA sets such limits, US law requires FDA either to set its own for bottled water or to find them unnecessary; FDA now says it "is working on a regulation to establish levels for certain PFAS in bottled water".

    Achieved whena final rule adding PFAS limits to 21 CFR 165.110 is published in the Federal Register.

    fda.govfda.gov

    Claimed by

    US Food and Drug Administration, page current as of July 2026

  2. Handling sheds plastic particles

    In a pilot on one brand, with one run per test, the testing laboratory Eurofins counted microplastic particles in water before and after everyday handling. Shaking a half-full bottle 100 times raised the count from 29 to 224 particles, and pressing the bottle 10 times raised it from 14 to 203. Two days in a hot car or in sunlight lowered the count, which Eurofins says may mean particles broke down below what its method can see.

    Achieved whena study with several brands and repeated runs, published with its method, finds the same effect.

    cdnmediaassets.eurofins.com

    Claimed by

    Eurofins, July 2026

Open

  1. 01

    Every brand's test report

    A free, public table of the published water quality report of every bottled water brand on sale in the United States, and later the EU and India, built by an agent that finds each report and pulls out each substance tested, its result, its unit and the legal limit. It logs brands that publish nothing or do not answer a written request, links every value to its source, and repeats the run each year.

    The brief
    Credit · unclaimed
  2. 02

    Labels rechecked since 2008

    A repeat, with open data, of the label check in the 2010 atlas of European bottled water: buy brands it tested, measure their main minerals at an accredited laboratory, and set each result beside the brand's current label and the 2008 measurement. Add brands from other continents, where label checks so far are small and national.

    The brief
    Credit · unclaimed

Open 01 · the brief

Every brand's test report

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A free, public table of the published water quality report of every bottled water brand on sale in the United States, and later the EU and India, built by an agent that finds each report and pulls out each substance tested, its result, its unit and the legal limit. It logs brands that publish nothing or do not answer a written request, links every value to its source, and repeats the run each year.

Why it is still open

GAO's 2009 recommendation that FDA find ways to give buyers this information is closed as not implemented. The last broad public survey, the Environmental Working Group's 2011 scorecard of 173 bottled waters bought in 2010, was done by hand; it found that 29 of 96 brands surveyed in California did not meet the state's disclosure law. Subscription phone apps now score brands from lab reports, but a May 2026 review of one, Oasis, found it had listed a laboratory's total strontium reading as radioactive strontium-90 and judged arsenic against a benchmark that is not a legal limit.

gao.govstatic.ewg.orgthedailypour.com

Built with what exists

A web-browsing agent such as Browser Use 1.0 (see the Note "Deleting 80% to start over"); a small, cheap AI model to pull figures out of PDF reports (see the Note "The jobs agents repeat"); the limits table in 21 CFR 165.110; Python with pdfplumber.

  1. First step

    List 20 brands from the water shelves of two supermarket chains, run the agent on them, and check every extracted value against the reports by hand.

  2. Done looks like

    Reports found or requested for at least 150 brands sold in the United States, each extracted value linked to its source page. 100 values drawn at random and checked by hand show no more than two errors. The table and code are public, and a second run a year later is published beside the first.

Open 02 · the brief

Labels rechecked since 2008

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A repeat, with open data, of the label check in the 2010 atlas of European bottled water: buy brands it tested, measure their main minerals at an accredited laboratory, and set each result beside the brand's current label and the 2008 measurement. Add brands from other continents, where label checks so far are small and national.

Why it is still open

The EuroGeoSurveys atlas, edited by Clemens Reimann and Manfred Birke, measured 1,785 bottled waters bought in 40 European countries in one laboratory and compared nine dissolved substances with the labels; reviewers report large differences for bicarbonate and silica. Its data came on a CD-ROM with the book. Outside Europe, a study of 21 Iranian brands bought in Isfahan in 2012 found measured calcium below the label value in 71.4% of brands. No repeat of the European check, and no open label check across continents, turned up in searches for this file.

schweizerbart.depmc.ncbi.nlm.nih.gov

Built with what exists

A phone camera; an open vision-language model fine-tuned to read label tables on one graphics chip (see the Note "One GPU to fine-tune"); ion chromatography and ICP-OES, an instrument that measures metals in a liquid, at a contract or university laboratory; the atlas's 2008 data; Python.

  1. First step

    Borrow the atlas and its CD-ROM from a library, check whether the data name each brand, list 20 named brands still on sale, and get quotes from two accredited laboratories for a major-ion panel.

  2. Done looks like

    60 brands, at least 40 of them tested for the atlas and at least 10 from outside Europe, two production lots each, with label photos, laboratory reports and the comparison table public. A tally, against a threshold of 20% fixed before testing, of brands whose measured major ions differ from the label and of atlas brands whose make-up has moved since 2008. A second laboratory reruns 10 of the samples and its results fall within the first laboratory's stated uncertainty.