Best Arsenic Water Filters (2026): What Actually Removes It

An under-sink reverse osmosis system certified to NSF/ANSI 58 with an explicit pentavalent arsenic reduction claim is the right choice for the large majority of households with arsenic in their water. It treats the water you actually drink and cook with, costs $150–$500, and installs in under an hour without a plumber in most kitchens.

Two caveats decide whether that recommendation holds for you. First, the certification must name arsenic — a system certified to NSF/ANSI 58 for TDS reduction alone has not been tested for arsenic at all. Second, if your water contains arsenite rather than arsenate, reverse osmosis alone will underperform badly. Both points are covered below, because both are routinely omitted from filter roundups.

A side benefit of the same membrane: reverse osmosis also removes nitrate, fluoride, and chromium-6.

TapWater.org does not test water filters in a laboratory, and we don't claim to. This guide is built from three sources we can point you at directly: the NSF certified-products database, which lists the specific reduction claims each system was actually tested for; EPA's Best Available Technology determinations for arsenic under the Safe Drinking Water Act; and the arsenic concentrations published in municipal Consumer Confidence Reports, which we aggregate across US cities.

Where we make a performance claim about a product, it comes from its certification listing rather than from marketing copy. Where we make a claim about arsenic chemistry, it comes from EPA's own rulemaking documentation. Anything we could not verify, we have left out.

Activated carbon removes contaminants by adsorption — organic molecules stick to the enormous surface area inside the carbon granules. Dissolved inorganic arsenic does not stick. It passes straight through, which is why a Brita pitcher, a PUR faucet mount and a refrigerator filter all leave arsenic concentrations essentially unchanged.

Water softeners are equally ineffective, for a different reason: ion-exchange resin in a softener is charged for calcium and magnesium, not for arsenic species. And boiling actively makes the problem worse. Water evaporates, arsenic does not, so the concentration in the remaining water rises.

EPA lists only three point-of-use or point-of-entry technologies as Best Available Technology for arsenic: point-of-use reverse osmosis, point-of-use activated alumina, and point-of-entry activated alumina. Everything else — anion exchange cartridges, pour-through devices, distillation units — may work in specific circumstances but has not met EPA's evidentiary bar for compliance purposes.

NSF/ANSI 58 is a system-level standard for point-of-use reverse osmosis, and it contains separate test protocols for separate claims. A manufacturer chooses which claims to certify. TDS reduction is the cheapest and most common. Pentavalent arsenic reduction is a distinct, additional claim that many otherwise-certified systems simply do not carry.

The practical consequence is that "NSF/ANSI 58 Certified" on a box tells you the system was tested against something in that standard. It does not tell you it was tested for arsenic. Two systems sitting next to each other on the same shelf, both displaying the same certification mark, can differ entirely on whether arsenic was ever measured.

There is a second layer to this. NSF/ANSI 58 is a whole-system certification. A membrane cartridge that is individually certified does not confer certification on the assembled unit it sits inside — a distinction manufacturers occasionally blur with phrasing like "built with NSF-certified components" or "tested to NSF standards."

Arsenic exists in drinking water in two inorganic forms, and reverse osmosis handles them very differently. Arsenate — arsenic V — carries a negative charge at normal drinking water pH and is rejected efficiently by an RO membrane. Arsenite — arsenic III — is electrically neutral under the same conditions, so the membrane has far less to grip. EPA states the position plainly in the arsenic rule: removal efficiencies for As(V) are much better than for As(III) across every technology evaluated, and where As(III) is present, pre-oxidation to As(V) is necessary.

This matters because of where arsenic comes from. As(V) dominates in aerobic surface water. As(III) dominates in anaerobic groundwater — which is exactly the source most likely to contain arsenic in the first place. If you are on a private well drawing from a reducing aquifer, arsenite is the form you should assume is present until a lab tells you otherwise.

Every NSF/ANSI 58 arsenic certification is written for pentavalent arsenic. None of them certify arsenite reduction. A system that performs exactly as certified can still leave you well above 10 ppb if your water is arsenite-dominated.

The fix is pre-oxidation, which converts As(III) to As(V) before the water reaches the membrane. Free chlorine, potassium permanganate and ozone are all effective oxidants — AquaTru's own performance documentation states that a free chlorine residual rapidly converts trivalent arsenic to pentavalent, and that ozone and potassium permanganate do the same. Chlorinated municipal water often does this conversion for you incidentally. Untreated well water generally does not, so a well system usually needs a chlorination or oxidation stage ahead of the RO — and aeration, despite being the intuitive choice, was found ineffective for arsenite oxidation in EPA-cited field work.

The chloramine wrinkle: the same documentation warns that combined chlorine — chloramine — may not convert all trivalent arsenic. Many US utilities disinfect with chloramine rather than free chlorine, so city water is only the more forgiving case if your utility uses free chlorine. Your utility's Consumer Confidence Report states which one it uses — ask, or check your city's page — before assuming the conversion has been done for you.

If your utility reports arsenic below roughly 2 ppb, an arsenic-specific filter solves a problem you don't have. Seattle, Portland, Fort Worth and Long Beach all sit in that band. A carbon filter for taste and chlorine byproducts is a more sensible purchase.

If you have not tested, do not buy yet. Arsenic is invisible, tasteless and odorless, and buying an RO system on suspicion is guesswork. Public water customers can read the utility's published value; private well owners need a lab test, and standard home test strips are not accurate at the concentrations that matter — a lab kit using EPA Method 200.8 detects down to 1 ppb. Our water test kit guide and well water test kit guide cover the options.

And if your concern is showering or bathing, you don't need arsenic treatment at all. Arsenic is not absorbed through skin in meaningful amounts during normal bathing at levels below 500 ppb. The exposure route is ingestion — drinking and cooking — which is precisely why point-of-use treatment at the kitchen tap is the correct architecture and a whole-house system is usually overspending.