This review was done before any figure was computed, because a study whose novelty is assumed gets dismissed on novelty. It found that the closest comparable work is a commercial certification-corpus study, not the academic literature — and that two premises this study was originally scoped on were wrong. Both corrections are carried below rather than buried.
VerifiedWaterFilterData (2026)
Evidence grade: read-in-full — retrieved and read in full.
Measured: the composition of the certified-filter corpus — which standard each certified filter holds, and how many filters are certified per contaminant.
Sample: 5,682 certified filters from NSF, WQA and IAPMO public listings.
Leaves unanswered: how many marketed claims have no listing behind them. Its universe is the certification listing, so a filter with no certification is absent from the denominator and the unbacked-claim population is unmeasurable by construction.
VerifiedWaterFilterData (July 2026). "Most Water Filters Are Certified for Taste — Not Health." Published under CC BY 4.0 with a downloadable dataset. Retrieved 2026-09-19.
MassDEP / UMass Amherst (2023)
Evidence grade: read-in-full — retrieved and read in full.
Measured: PFAS6 in the effluent of consumer point-of-use devices at 100% and 200% of the manufacturer's rated throughput, against the Massachusetts PFAS6 MCL of 20 ng/L.
Sample: 17 devices, 51 water samples, one water source — the Harbor Trace well, Townsend, Massachusetts.
Leaves unanswered: whether the contaminants these devices advertise removing are listed anywhere. It used manufacturer claims to select devices and then measured performance, never scoring one against the other.
MassDEP & UMass Amherst Department of Civil and Environmental Engineering, Per- and Polyfluoroalkyl Substances (PFAS) Point-of-Use Treatment System Study, Final Report, November 2023. 12 of the 17 devices exceeded the MCL at rated volume; 3 of the 17 held NSF/ANSI 53 certification and all three passed.
Tang, Lytle, Achtemeier & Tully (2023)
Evidence grade: read-in-full — retrieved and read in full.
Measured: whether NSF/ANSI 53 certified filters achieve the certification benchmark for lead in practice.
Sample: 1,678 filters across 23 studies — the largest device sample in the literature.
Leaves unanswered: whether a filter's advertising matches its listing. Its population is defined by already holding the certification, so an uncertified over-claimer cannot enter it.
Tang, M., Lytle, D., Achtemeier, R., Tully, J. (2023). "Reviewing performance of NSF/ANSI 53 certified water filters for lead removal." Water Research 244:120425. 96% of the tested devices met the benchmark.
Teymoorian, Dinh, Barbeau & Sauvé (2024)
Evidence grade: read-in-full — retrieved and read in full.
Measured: removal of up to 75 targeted PFAS by pitcher-type point-of-use filters across two contrasting tap-water sources.
Sample: 5 pitcher filters, 2 water sources.
Leaves unanswered: how often the pairing disagrees across the market. Its Table 1 puts the held certificate next to the marketed PFOA/PFOS claim for five filters and stops there.
Teymoorian, T., Dinh, Q.T., Barbeau, B., Sauvé, S. (2024). "Performance of pitcher-type POU filters for the removal of 75 PFAS from drinking water." Frontiers in Environmental Chemistry. It records that Aquagear is not NSF-certified for PFAS removal and that the manufacturer provides third-party performance data on its website.
Herkert et al. (2020)
Evidence grade: secondary — not read in full — figures taken from two independent secondaries that agree.
Measured: PFAS removal by in-home filters as actually installed and maintained by residents.
Sample: 89 systems across 73 residences in North Carolina.
Leaves unanswered: what any of those systems was certified or advertised to remove — certification status is not recorded at all.
Herkert, N.J. et al. (2020). Environmental Science & Technology Letters 7(3):178-184. The publisher returned HTTP 403 to every transport tried, so the figures here come from two independent secondaries that agree: Duke University's press release and the description in Teymoorian et al. (2024).
Consumer Reports
Evidence grade: unretrieved — not retrieved (paywalled).
Measured: taste and odour reduction, flow rate and clogging in its own laboratory; for contaminants it verifies the manufacturer's certification claim rather than measuring it.
Sample: not disclosed — the ratings are behind a paywall.
Leaves unanswered: any aggregate. It performs this study's check one model at a time and publishes no count of claims checked, no count that failed, and no dataset.
Consumer Reports, Best Water Filter Buying Guide, section "How CR Tests Water Filters". The public methodology page was read in full; the ratings tables and any sample size are paywalled and were not retrieved.
European Commission / CPC Network (2021)
Evidence grade: read-in-full — retrieved and read in full.
Measured: whether online environmental marketing claims were substantiated.
Sample: 344 claims examined in detail, screened by consumer authorities across 27 countries.
Leaves unanswered: the same audit in a sector with a public certification register to score against, which is what makes a verdict checkable rather than a matter of judgement.
European Commission, IP/21/269, 28 January 2021, "Screening of websites for greenwashing: half of green claims lack evidence." In 42% of cases authorities had reason to believe the claim may be false or deceptive — note the wording the regulator chose.
WQA Gold Seal programme guide
Evidence grade: read-in-full — retrieved and read in full.
Measured: nothing — it is the scheme document.
Sample: not applicable.
Leaves unanswered: the distance between the literature a manufacturer submits and what it then says at retail. WQA's literature review evaluates the installation manual, the performance data sheet and the data label; none of them is the product page.
Water Quality Association, The Gold Standard: Your Guide to WQA's Gold Seal Product Certification Program.
US EPA consumer guidance
Evidence grade: read-in-full — retrieved and read in full.
Measured: nothing — it is consumer guidance.
Sample: not applicable.
Leaves unanswered: how often the check it recommends would fail. EPA tells a consumer to verify the listing themselves and names the accredited bodies to verify it with.
US EPA, Identifying Drinking Water Filters Certified to Reduce PFAS, last updated 18 November 2025.
US FTC Health Products Compliance Guidance
Evidence grade: read-in-full — retrieved and read in full.
Measured: nothing — it is enforcement guidance.
Sample: not applicable.
Leaves unanswered: how many filter claims in market meet the bar. It requires competent and reliable scientific evidence before a claim is made, and does not treat third-party certification as the only way to hold it.
US FTC, Health Products Compliance Guidance, section III.B. The asymmetry it creates is load-bearing for this study: an uncertified claim is not automatically an unsubstantiated one.
One correction is worth stating on its own, because the loose version of it circulates. The Massachusetts point-of-use study is often summarised as showing that filters fail to meet "state contaminant standards". It tested one family of contaminants against one standard, on influent from one well in Townsend, Massachusetts — and the devices in it that held NSF/ANSI 53 certification all passed. It is evidence that certification worked, not that it failed, and anyone pitching it the other way will be corrected by the people most likely to check.