Do water filters work? What certification proves and what it means for your health

What federal rules require of your water utility

Certification means the model passed a standard test for a contaminant, and NSF retests certified products periodically and inspects factories every year^{7, 9}. EPA’s own advice points people who use a filter to certified models for lead and PFAS^{6, 11, 13}.

In a blinded Australian trial on chlorinated city water, real filters didn’t reduce stomach illness compared with sham units¹⁶. We found no trial showing that a home filter improves health in people whose water already meets federal limits. Removing PFOA from water does lower blood levels, but slowly, by about a quarter a year¹⁷.

In April 2024, EPA finalized national drinking water limits for six PFAS¹. PFOA and PFOS are each capped at 4.0 parts per trillion (ppt)¹. PFHxS, PFNA and HFPO-DA (often called GenX) are capped at 10 ppt each¹. A separate hazard index limits mixtures of PFHxS, PFNA, GenX and PFBS¹. Utilities have until 2027 to finish their first round of testing and until 2029 to fix any exceedances¹.

Those deadlines may move. On May 18, 2026, EPA proposed keeping the PFOA and PFOS limits but letting water systems request two more years, to 2031². Under that proposal, systems with a PFOA or PFOS result at or above 12 ppt would have to take short-term mitigation steps².

A second proposal would rescind the limits for PFHxS, PFNA, GenX and the mixture³. Comment periods for both closed on July 20, 2026^{2, 3}. EPA’s pages for both, last updated October 8, 2026, still call them proposed rules^{2, 3}.

Lead falls under a separate rule, the Lead and Copper Rule Improvements (LCRI). It took effect on December 30, 2024, and lowers the lead action level from 15 to 10 micrograms per liter⁴. Water systems must start complying three years after the rule was issued, in late 2027⁴.

Most must then replace all lead and certain galvanized service lines under their control within 10 years⁴. Systems with continually high lead levels must make filters certified to reduce lead available to all their customers⁴. At homes with lead service lines, systems must take first-liter and fifth-liter tap samples and use the higher result⁴.

EPA’s LCRI page, last updated December 29, 2025, still presents the rule as final, with lead pipes replaced within 10 years⁵. A federal limit tells you what your utility must deliver, not what comes out of your tap. Lead usually gets in on the way, from lead service lines and plumbing⁶. Lead pipes are more likely in older cities and in homes built before 1986⁶. EPA’s goal for lead in drinking water is zero, and EPA and CDC say there’s no known safe level of lead in a child’s blood⁶.

What NSF/ANSI 42, 53, 58, 401 and 473 certify

A certification covers specific contaminants, not your water in general. NSF says certification doesn’t mean a filter reduces every possible contaminant, so check the claim for the one you’re worried about⁷. NSF isn’t the only certifier. EPA lists five accredited certifiers, each with its own online product listing: CSA Group, IAPMO R&T, NSF, UL and WQA¹¹.

NSF says certified products are retested periodically and their factories inspected every year⁹. A maker’s own lab report doesn’t come with that follow-up. Here’s what each standard named on filter boxes covers.

• NSF/ANSI 42 covers aesthetic impurities such as chlorine, taste and odor⁷. It isn’t a health claim. Chloramine claims also sit under it⁸. EPA’s lead guide also recommends a 42 claim for Class I particulate reduction¹³.
• NSF/ANSI 53 covers contaminants with a health effect, as regulated by EPA and Health Canada⁷. NSF’s claims guide lists lead, PFOA and PFOS, VOCs and trihalomethanes among them⁸. Each contaminant is a separate claim, so a filter certified for VOCs isn’t necessarily certified for lead.
• NSF/ANSI 58 covers reverse osmosis (RO) systems that reduce contaminants regulated by EPA and Health Canada⁷. On NSF’s list, fluoride and nitrate claims appear only on RO systems⁸.
• NSF/ANSI 401 covers emerging contaminants, such as some pharmaceuticals and chemicals EPA and Health Canada don’t yet regulate⁷. A filter can be certified for one or more of 15 emerging contaminants, so check which ones⁷.
• NSF/ANSI 473 isn’t described on NSF’s consumer guide to filter standards, and a search of NSF’s listings under 473 returned no products^{7, 10}. NSF says PFAS reduction claims are certified under 53 and 58⁹. EPA’s home filter advice names only those two¹¹. If a box cites 473, look up the certifier’s listing to see which standard and contaminants it actually covers.

Do PFAS-certified filters meet EPA’s new limits?

NSF says a filter must reduce PFAS to below 20 ppt to make a PFAS reduction claim⁹. That’s five times EPA’s 4.0 ppt limit for PFOA or PFOS¹. EPA’s home filter fact sheet said that, as of April 2024, certifications didn’t yet show a filter will remove PFAS down to its new limits¹¹. NSF says its committees are proposing updates to match EPA’s rule, but it gives no timeline⁹.

The same gap shapes EPA’s rules for utilities. EPA hasn’t designated point-of-use or whole-house (point-of-entry) units as a way for water systems to comply, because the PFAS limits sit below current certification levels¹². It expects that could change for small systems once certifications are revised¹². None of this means a certified filter doesn’t work. It means the label promises less than the new federal limits ask of a utility.

What does reverse osmosis remove? Minerals too

Reverse osmosis forces water through a very thin barrier that separates chemicals from the water¹¹. On NSF’s list, it’s the only filter type with certified claims for fluoride and for nitrate and nitrite⁸. RO systems also carry claims for lead, arsenic, hexavalent chromium and PFOA and PFOS⁸.

RO doesn’t sort useful minerals from contaminants. iSpring, for example, says its RO500AK-BN removes up to 99% of over 1,000 contaminants and lists calcium and sodium among them¹⁴. Its third stage adds minerals back to balance pH and improve taste¹⁴. iSpring doesn’t say which minerals or how much¹⁴. It also says the system is NSF 58 certified for reducing total dissolved solids (TDS)¹⁴.

Whether losing water’s minerals matters for health is unsettled. A 2016 meta-analysis pooled 10 studies with 77,821 coronary heart disease cases, nine of them from Europe¹⁵. Higher magnesium in drinking water was linked with lower coronary heart disease death (relative risk 0.89, 95% CI 0.79 to 0.99)¹⁵.

Only three were cohort studies, and on their own they showed no significant link¹⁵. Results also varied widely between studies¹⁵. These studies compared local water supplies, not people using RO at home.

Does a certified filter remove what it claims?

For the contaminants on its listing, yes, under the conditions it was tested in. That’s what certification checks, and NSF follows up with periodic retesting and yearly factory inspections⁹. EPA leans on that system in its own advice. For lead, it tells you to make sure a filter is certified to remove lead⁶. For PFAS, it says to choose a filter certified under NSF/ANSI 53 or 58¹¹.

EPA names three technologies that can reduce PFAS at home: activated carbon, reverse osmosis and ion exchange resins¹¹. Its lead guide says to look on the filter or its box for a mark from an accredited certifier and a lead reduction claim under NSF/ANSI 53¹³. If the claim isn’t on the box, EPA says to check the performance data sheet, the maker’s website or the certifier’s online listing¹³.

Upkeep matters as much as the label. EPA says using a cartridge after it expires can make it less effective at removing lead⁶. Its PFAS fact sheet says filters only work if they’re maintained as the maker instructs¹¹. EPA also suggests checking whether PFAS are in your water first. If levels are below federal limits, it says a filter may not be useful to you¹¹.

Does filtered water improve health?

In a blinded trial in Melbourne, Australia, 600 families got either a real water treatment unit or a sham one¹⁶. The real units were built to remove viruses, bacteria and protozoa¹⁶. Families kept weekly health diaries for 68 weeks¹⁶.

The city’s water came from a protected catchment and was treated with chlorine only¹⁶. Rates of gastroenteritis were about the same in both groups (rate ratio 0.99, 95% CI 0.85 to 1.15)¹⁶. The authors found no evidence of waterborne illness in Melbourne¹⁶.

For PFAS, the clearest evidence is about blood levels, not illness. In Ohio and West Virginia, two public water systems contaminated with PFOA added granular activated carbon filtration in 2007¹⁷. Among 200 adults, blood PFOA then fell by an average of 26% a year (95% CI 25% to 28%)¹⁷. The median half-life was 2.3 years¹⁷.

People on public water who used a carbon filter at home weren’t eligible, so this reflects treatment at the utility, not home filters¹⁷. It shows that cutting PFOA in water lowers blood levels slowly. It doesn’t show how much that changes anyone’s health.

For lead, the case for a filter rests on avoiding exposure rather than on filter trials. EPA says even low levels of lead in a child’s blood can cause behavior and learning problems, lower IQ and hearing problems⁶. In adults, it lists effects such as higher blood pressure and reduced kidney function⁶. None of the sources reviewed here measured blood lead in people before and after they started using a home filter.

What the evidence doesn’t show

Certification is good evidence that a filter cuts a named contaminant. It says much less about health, especially when your water already meets federal limits.

• A claim covers only the contaminants on the listing⁷. A cartridge used past its replacement date can also be less effective, at least for lead⁶.
• PFAS certification lags the new limits. A PFAS claim needs results below 20 ppt, while EPA’s limits for PFOA and PFOS are 4.0 ppt^{1, 9}. EPA said in 2024 that certifications didn’t yet show removal to those levels¹¹.
• Health trials of home filters are scarce. The blinded Melbourne trial, on chlorinated city water, found no drop in stomach illness¹⁶. It doesn’t tell you about water with a known lead or PFAS problem.
• The blood PFOA decline came from carbon filtration at the water plant, not home filters¹⁷. The study didn’t test whether a home filter does the same.
• The mineral evidence is observational. The magnesium findings come mostly from European studies of local water supplies, with wide variation between them¹⁵.
• None of the sources reviewed here tested claims that filtered water improves energy, hydration or longevity.

What to look for when buying a home filter

Start with your water, not the filter. Your utility’s test results, or a lab test for a private well, tell you what needs removing. EPA makes the same point for PFAS: if they’re not in your water above federal limits, a filter may not be useful¹¹. Then pick a filter certified for exactly those contaminants.

• Look up the listing. Search the certifier’s online directory for the exact model, and check which standard and contaminants it covers^{11, 13}. The listing confirms that the mark on the box applies to that model.
• For lead, look for a lead reduction claim under NSF/ANSI 53 and a Class I particulate claim under NSF/ANSI 42¹³. Use only cold water for drinking, cooking and baby formula⁶. Boiling doesn’t remove lead⁶.
• For PFAS, look for a PFOA and PFOS claim under NSF/ANSI 53 or 58¹¹. A certified claim only promises results below 20 ppt in testing, not below EPA’s 4.0 ppt limits^{1, 9}.
• For fluoride or nitrate, only RO systems carry certified claims on NSF’s list⁸. Expect RO to take out minerals too. Some units add a remineralizing stage, but makers may not say what it adds¹⁴.
• For chlorine taste and smell, an NSF/ANSI 42 claim is enough. Don’t read it as a health claim⁷.
• Budget for cartridges and change them on time. An expired cartridge can be less effective against lead⁶. EPA says a filter can cost from $20 to more than $1,000, not counting upkeep¹¹.

Sources

1. Per- and Polyfluoroalkyl Substances (PFAS): PFAS National Primary Drinking Water Regulation US Environmental Protection Agency, 2026 · regulator
2. Proposed PFOA and PFOS Compliance Extension Rule US Environmental Protection Agency, 2026 · regulator
3. Proposed PFAS Rescission Rule US Environmental Protection Agency, 2026 · regulator
4. National Primary Drinking Water Regulations for Lead and Copper: Improvements (LCRI), final rule (89 FR 86418) Federal Register (US Environmental Protection Agency), 2024 · regulator
5. Lead and Copper Rule Improvements US Environmental Protection Agency, 2025 · regulator
6. Basic Information about Lead in Drinking Water US Environmental Protection Agency, 2026 · regulator
7. NSF Standards for Water Treatment Systems NSF · certifier
8. Contaminant Reduction Claims Guide NSF · certifier
9. Perfluorooctanoic Acid (PFOA) and Perfluorooctanesulfonic Acid (PFOS) in Drinking Water NSF · certifier
10. NSF listings: drinking water treatment units, Standard 473 (no matching products found) NSF · certifier
11. Reducing PFAS in Your Drinking Water with a Home Filter (fact sheet) US Environmental Protection Agency, 2024 · regulator
12. Treatment Options for Removing PFAS from Drinking Water (fact sheet) US Environmental Protection Agency, 2024 · regulator
13. A Consumer Tool for Identifying Point of Use (POU) Drinking Water Filters Certified to Reduce Lead (EPA/600/F-18/335) US Environmental Protection Agency, 2018 · regulator
14. iSpring RO500AK-BN NSF Certified Tankless Reverse Osmosis Water Filtration System iSpring · manufacturer
15. Magnesium Levels in Drinking Water and Coronary Heart Disease Mortality Risk: A Meta-Analysis Nutrients (Jiang L, He P, Chen J, et al.), 2016 · meta-analysis
16. A randomized, blinded, controlled trial investigating the gastrointestinal health effects of drinking water quality Environmental Health Perspectives (Hellard ME, Sinclair MI, Forbes AB, Fairley CK), 2001 · randomized trial
17. Rate of Decline in Serum PFOA Concentrations after Granular Activated Carbon Filtration at Two Public Water Systems in Ohio and West Virginia Environmental Health Perspectives · cohort study