The numbers from the only large independent comparison test we have are not subtle. Under-sink reverse osmosis approaches complete PFAS removal in finished water; most carbon-pitcher filters perform between roughly zero and 30 percent reduction, and a handful are worse than no filter at all because they concentrate short-chain PFAS into the drinking glass.
Per- and polyfluoroalkyl substances are not like chlorine, lead, or sediment. They are short, very polar molecules with a fully fluorinated carbon backbone. The carbon–fluorine bond is the strongest single bond in organic chemistry, which is the entire reason these compounds persist in the environment and in human serum for years. Filtration has to physically separate the PFAS molecule from the water; nothing in a typical pitcher chemically breaks it down.
Granular activated carbon, the medium inside almost every Brita, PUR, and supermarket-brand pitcher, works by adsorbing organic contaminants onto the surface of carbon granules. For long-chain PFAS like PFOA (eight carbons) and PFOS (eight carbons with a sulfonate head), adsorption to GAC is reasonably effective when contact time is long enough and the carbon bed is fresh. Municipal-scale GAC contactors achieve high PFOA/PFOS removal because water sits in the bed for minutes and the bed is monitored for breakthrough. A pitcher passes water through a few millimetres of carbon in seconds.
The short-chain PFAS, PFBA, PFBS, PFHxA, and EPA's regulated replacement compound HFPO-DA (GenX), adsorb far more weakly to GAC. They are smaller, more soluble, and out-compete long-chain PFAS for adsorption sites. A pitcher cartridge that did a respectable job on PFOA in laboratory water will routinely fail on the GenX in Cape Fear basin tap water. EPA's own treatment guidance (Reducing PFAS in Drinking Water with Treatment Technologies) is explicit that GAC performance varies "significantly across PFAS compounds, with greater removal of long-chain PFAS than short-chain PFAS."
That is the chemistry baseline. The rest of this guide is the field test.
The most-cited independent comparison of point-of-use PFAS filters is Herkert et al. (2020), "Assessing the Effectiveness of Point-of-Use Residential Drinking Water Filters for Perfluoroalkyl Substances (PFASs)," published in Environmental Science & Technology Letters 7(3), 178–184 (DOI: 10.1021/acs.estlett.0c00004). The group sampled influent and effluent from filters installed in actual homes, 76 systems across North Carolina including pitcher filters, refrigerator filters, faucet-mounted filters, two-stage under-sink units, and reverse osmosis. Analysis was by EPA Method 537.1 for the standard suite of long- and short-chain PFAS.
The headline finding the paper reports verbatim: under-sink reverse osmosis and two-stage filters achieved "near-complete removal of all PFAS." Activated carbon filters in pitcher, faucet, and refrigerator formats achieved "highly variable" removal, with some filters showing PFAS concentrations increasing after treatment, meaning a cartridge past its useful life can release previously adsorbed PFAS back into the water passing through it.
The averaged removal rates from the paper, summarised:
| Filter type tested | Mean PFAS reduction | Notes from Herkert et al. 2020 |
|---|---|---|
| Under-sink reverse osmosis | ~94% (approaching 100%) | Reduced all measured PFAS to near or below the analytical reporting limit. |
| Two-stage under-sink (carbon + secondary) | ~73% | Strong on long-chain PFAS; weaker but still substantial on short-chain. |
| Activated-carbon pitcher / faucet / fridge | Highly variable; near 0% in worst cases | Several units showed negative reduction (effluent > influent) consistent with cartridge breakthrough. |
| Whole-house carbon (POE) | Variable, often poor | Contact time and bed exhaustion limit residential POE units; not recommended as the sole barrier. |
Two qualifications matter when reading these numbers. First, the percentages are averaged across the measured PFAS suite, individual compounds vary, and short-chain PFAS specifically pull pitcher and faucet averages down. Second, the home-installed units were of varying age and cartridge state, which is exactly the real-world condition we care about. A brand-new pitcher in a laboratory will outperform the average; a six-months-overdue pitcher under a real kitchen sink can hit zero.
The marketing copy on filter boxes is engineered to imply PFAS protection without claiming it. The only consumer-side claims that mean anything specific about PFAS are these:
Anything else on the box, "NSF 42 certified," "tested to NSF standards," "uses an NSF-component-tested filter," "filters 99% of contaminants", does not mean the filter has been certified to reduce PFOA or PFOS. NSF 42 covers aesthetic effects (taste, odour, chlorine). The "component tested" language indicates that some part of the unit passes a materials-safety test, not that the assembly removes PFAS.
Our companion guide, NSF 53 P473 vs NSF 58, which certification actually covers PFAS, walks through the difference in test protocols and what to look for on a NSF listing page.
A residential reverse osmosis unit forces tap water through a semipermeable thin-film composite (TFC) polyamide membrane under household line pressure (typically 40–80 psi). The membrane pore size is on the order of 0.0001 micrometres, which is small enough to reject the hydrated PFAS anion regardless of chain length. The clean permeate goes to the dedicated tap; the concentrate (containing the rejected PFAS, dissolved solids, and most other contaminants) goes to the drain. Modern units waste between 1 and 3 gallons per gallon produced.
The Duke 2020 measurements showed RO units approaching 100% reduction across the full PFAS suite they tested, including the short-chain compounds that defeat GAC. This is consistent with bench-scale literature on TFC polyamide membranes, which typically report 95–99% rejection of individual PFAS species. A well-installed under-sink RO with NSF 58 certification and a PFOA/PFOS reduction claim is the closest thing to a definitive answer for a residential customer on impacted source water.
Practical notes on RO. First, the membrane itself is the workhorse, but the system also includes a sediment pre-filter and a carbon pre-filter that protect the membrane from chlorine. Skipping the recommended pre-filter change schedule will degrade the membrane and pull performance down over time. Second, the storage tank is metal or plastic with a bladder; smell and taste issues usually trace to a stale tank rather than the membrane. Third, the remineralisation cartridge (sold as a "alkaline" or "pH" stage) is optional and does not affect PFAS rejection.
If your utility appears in our PFAS groundwater hotspot map, Cape Fear, Wolverine in Michigan, Hoosic and Hudson valleys in NY/VT, Decatur in Alabama, the Colorado Front Range military airbases, under-sink RO is the recommendation, not carbon-block. The short-chain PFAS associated with these hotspots are exactly what carbon misses.
Plumbed under-sink RO needs a sink drain connection and a hole in the counter for the dedicated tap. That is a non-starter in most rentals. The countertop RO category, AquaTru, Bluevua, Waterdrop A1, and a handful of newer entrants, solves this by packaging the membrane, pumps, and storage into a benchtop appliance. You pour tap water in, it pressurises against the membrane, the permeate fills a serving tank, and the concentrate goes to a waste tank you empty by hand.
Independent testing on countertop RO units is sparser than for under-sink, but the technology is the same TFC polyamide membrane in a smaller housing, and the manufacturer NSF 58 certifications generally hold up. AquaTru in particular has been NSF-certified for PFOA and PFOS reduction. Throughput is lower (typically 1 gallon per 10–15 minutes) and waste ratios are similar to plumbed units. The premium over plumbed RO is real, expect $400–600 upfront and $120–200 a year in cartridges versus $200–700 / $50–120 for an installed system.
For renters in a hotspot ZIP, a countertop RO is the right answer. For owners, plumb the under-sink unit.
The Duke 2020 paper found one notable exception inside the pitcher category: two-stage units that pair GAC with an ion-exchange or secondary-media stage performed materially better than single-stage carbon pitchers. ZeroWater's five-stage ion-exchange pitcher and the Travel Berkey gravity unit fall in this group. The paper measured reductions in the 70%+ range for these formats on the PFAS species they detected, not RO-class numbers, but a meaningful step up from a Brita.
The caveats are real. ZeroWater cartridges exhaust quickly, especially in high-TDS water, and a spent cartridge can release captured PFAS. Berkey gravity units are unregulated as a category and have been the subject of EPA enforcement actions over uncertified pesticide claims; the company's own PFAS testing is third-party but not NSF-certified, which puts it outside our normal evidence bar. We mention both because the peer-reviewed data is what it is: in the Duke samples, these formats outperformed standard carbon pitchers. They are not equivalent to RO and we would not recommend either over a NSF 58 under-sink unit for a household on hotspot water.
The sticker price comparison favours pitchers; the three-year total cost is closer than people expect, and the cost-per-litre-of-actually-clean-water is dramatically in RO's favour because much of the pitcher output is not meaningfully PFAS-reduced.
| System | Upfront | Annual cartridges | 3-year total | PFAS reduction (Duke 2020) |
|---|---|---|---|---|
| Under-sink RO (NSF 58 + P473) | $300 | $85 | $555 | ~94% (near 100%) |
| Countertop RO (AquaTru class) | $500 | $160 | $980 | ~94% (near 100%) |
| Two-stage under-sink (carbon) | $180 | $90 | $450 | ~73% average |
| ZeroWater pitcher | $40 | $120 | $400 | ~70% (variable) |
| Standard carbon pitcher | $30 | $50 | $180 | 0–30% (often zero) |
The under-sink RO line is cheaper over three years than a countertop and within $100 of a standard pitcher over the same period, while delivering an order of magnitude more PFAS reduction. The standard pitcher is the only entry where the dollar saving is real, and the trade is a filter that does essentially nothing for the contaminant you bought it to remove.
If you don't yet know what is in your tap, that's the first thing to settle. Our home PFAS testing labs guide covers the four certified labs that will run an EPA Method 537.1 or 533 panel on a mailed-in sample for around $300. If you want utility-side context first, the UCMR 5 explainer and the April 2024 MCL summary are the two priors that frame everything else.
State UCMR-5 baseline, hotspot utility match if applicable, the 6 regulated PFAS with their MCLs, a filtration matrix with 2024 prices and certified models, lab options to test your own tap, and a 14-day plan. Instant PDF, no subscription, sources cited.
Last reviewed: 30 June 2026 · How Safe Is My Water editorial. Independent. Not affiliated with the US EPA, NSF International, or any filter manufacturer.
