How Often to Change Every Type of Water Filter
Six months is a marketing number, not an engineering one. Here is what each component’s own manufacturer says, and what makes the real answer shorter.
Compiled under our editorial and verification guidelines. No manufacturer paid for inclusion.
How often you change a water filter depends on the filter type, its rated gallon capacity, how much water your household actually puts through it, the sediment and contaminant load, whether you are on a well or a main, any change in flow or pressure, and the manufacturer’s own instructions — which outrank every general rule including this page. Real intervals run from two months to ten years depending on the component. A pitcher cartridge and a softener resin bed are both “filters”, and one lasts sixty times longer than the other.
Replace by the earliest applicable limit: elapsed time, rated capacity, a pressure-drop threshold, a test result, or a manufacturer requirement. Whichever comes first governs — not whichever is most convenient. A 10,000-gallon under-sink cartridge in a two-person home will hit its twelve-month calendar limit long before it hits its capacity rating, so the calendar wins. A whole-house cartridge on a silty well can exhaust its capacity in eight weeks, so capacity wins. And any of them can be overridden by a failed bacteria test, which is why a filter is never the answer to a contamination question you have not asked a laboratory.
Why is there no universal replacement schedule?
Because the number that actually governs is gallons, and gallons are a household variable rather than a product one. A cartridge rated for 300 gallons meets its rating in a hundred days at three gallons a day, or fifty days at six. Same cartridge, half the life.
Three other things move it. Sediment load physically blinds a cartridge — a well with visible grit can end a 100,000-gallon whole-house cartridge months early, and no rating survives that. Water chemistry shortens some components specifically: chlorine degrades softener resin, and iron fouls it. Standing time matters too; a filter that sits unused is not paused, because the media stays wet.
EPA’s consumer guidance on filter maintenance is unambiguous: follow the instructions that came with your filter, because each type, brand and model differs, and change the cartridge on time because cartridges “wear out and become less effective with use” Regulatory guidance. Every interval on this page is subordinate to the manual in your drawer. Where our figure and the manufacturer’s disagree, theirs governs — they tested the specific media, and we did not.
Master replacement-frequency table
| Component | Typical starting interval | Rated-capacity method | Signs of early exhaustion | What changes it | Best trigger | Source |
|---|---|---|---|---|---|---|
| Pitcher cartridge | 2–4 months | 40–150 gal typical; ZeroWater states an estimated 15–20 gal | Slow fill, taste returns, TDS meter rises | Household size; feed-water TDS on ion-exchange types | Gallons, then calendar | Manufacturer Clearly Filtered 100 gal; ZeroWater 15–20 gal |
| Faucet-mounted | 2–3 months | ~100 gal | Indicator light; flow drops sharply | Heavy cooking use; sediment | Indicator, then gallons | EPA/Flint programme ~100 gal or when the indicator turns red |
| Refrigerator filter | 6 months | 200–300 gal typical | Ice tastes of chlorine; slow dispenser | Ice-maker use, which owners routinely underestimate | Calendar, then gallons | Manufacturer major-brand OEM filters |
| Countertop gravity | 2–6 months | 100–150 gal | Fill time lengthens noticeably | Household size; sediment | Gallons | Manufacturer Epic Pure XP 3–4/yr ⚠ capacity stated three ways on one page |
| Countertop RO | 6 mo to 2 yr by stage | Pre 600 gal · VOC 600 gal · membrane 1,200 gal | TDS readout climbs; slower batch | Feed TDS; hardness scaling the membrane | Whichever stage comes due | Manufacturer AquaTru Classic |
| Under-sink carbon | 6–12 months | Up to 10,000 gal on large blocks | Chlorine taste returns; visible flow loss | Sediment upstream | Calendar — capacity rarely binds first | Manufacturer typical carbon-block ratings |
| RO sediment & carbon pre-filters | 6–12 months | Varies; Waterdrop states CF 550 gal, CB 1,100 gal | Production rate falls; membrane fails early | Sediment. These protect the membrane — skipping them is expensive | Calendar | Manufacturer APEC 6–12 mo; Waterdrop 6/12 mo |
| RO membrane | 2–4 years | 1,200 gal (countertop) to several thousand | TDS rejection falls; tank fills slowly | Feed TDS, hardness, chlorine reaching the membrane | TDS test, then calendar | Manufacturer APEC 2–4 yr; AquaTru 2 yr; Waterdrop 24 mo |
| RO post-filter | 1–4 years | Polishing stage, low load | Stale or flat taste from the tank | How long water sits in the tank | Calendar | Manufacturer APEC 2–4 yr |
| Whole-house cartridge | 6–12 months | 100,000 gal for a 20″ Big Blue stage | Measurable pressure drop across the housing | Sediment load, more than anything else | Pressure drop, then calendar | Manufacturer iSpring WGB32B 100,000 gal / 12 mo |
| Spin-down sediment | Never — flush it | Reusable stainless screen | Visible load in the bowl; pressure drop | Well silt, mains work, hydrant flushing | Visual inspection, monthly | Manufacturer reusable-screen designs |
| Tank-based carbon media | 5–10 years | Media volume × capacity, not a cartridge count | Chlorine breakthrough at the tap | Chloramine shortens it; contact time governs | Test for breakthrough | Editorial trade convergence |
| Iron / manganese media | 3–7 years, plus regeneration | Oxidising media, regenerated not replaced | Staining returns between backwashes | Iron form and concentration; pH | Return of staining, then test | Editorial trade convergence |
| Softener resin | 10–15 years | Not a cartridge — a bed that degrades | Hardness returns despite adequate salt | ⚠ Chlorine oxidises it; iron fouls it. Both can halve its life | Hardness test | Editorial trade convergence, no agency figure exists |
| Softener prefilter | 6–12 months | As for any sediment cartridge | Pressure drop | Sediment | Calendar | Editorial |
| UV lamp | 12 months | Rated ~9,000 hours ≈ 375 days of continuous run | None visible — output decays invisibly | Nothing extends it. Power cycling shortens it | Calendar, without exception | Manufacturer UV installation manuals |
| UV quartz sleeve | 2–3 years | Clean at each lamp change | Visible fouling or haze through the port | Hardness above ~7 gpg; iron | Inspection at lamp change | Manufacturer manuals advise softening above 7 gpg |
| UV prefilter | 6–12 months | Sediment rating | Pressure drop | ⚠ Critical — turbidity shields organisms from the lamp | Calendar | Manufacturer |
| Shower filter | 4–6 months | ~10,000 gal typical | Chlorine smell returns in steam | Shower frequency and length; hot water degrades media faster | Calendar | Manufacturer |
[Master frequency table] Ranges, not promises. Manufacturer rows cite a specific maker’s published interval; Editorial rows are trade convergence where no agency or certifier figure exists, and we flag them rather than borrowing false authority. Your manual overrides any row here.
What changes for each type of filter?
Point-of-use filters run out of gallons. Pitchers, faucet-mounts and refrigerator cartridges hold very little media, so capacity is almost always the binding limit — and the number people underestimate is the ice-maker, which quietly doubles a fridge filter’s throughput. Ion-exchange pitchers are the outlier: their life is set by how many dissolved solids the feed water carries, not by a gallon count, so the same cartridge can last weeks or months in different houses. Under-sink carbon blocks are the opposite case, holding enough media that the calendar cap almost always arrives first.
Reverse osmosis is four clocks, not one. The sediment and carbon pre-filters exist to protect the membrane, and neglecting them is how a two-to-four-year membrane dies in eighteen months. Chlorine reaching a thin-film membrane damages it, which is precisely what the carbon pre-filter prevents. The post-filter polishes water that has been sitting in the tank and carries almost no load, so it runs longest.
Whole-house and well components fail on load and chemistry rather than time. A cartridge stage is governed by sediment; a spin-down screen is not replaced at all, only flushed, and inspecting it monthly is free. Tank-based carbon is measured in years because contact time does the work rather than a cartridge. Iron and manganese media regenerate rather than exhaust, so the signal is staining returning between backwashes. Softener resin degrades chemically — chlorine oxidises it and iron fouls it, and either can halve a ten-to-fifteen-year bed. UV is the one component with no discretion at all: the lamp is rated at roughly 9,000 hours and its output decays invisibly, so it is replaced annually whether or not it still glows.
How do you calculate replacement timing from rated gallons?
Estimated replacement days = rated gallon capacity ÷ household gallons filtered per day
Then cap the result at the manufacturer’s maximum time interval, and take whichever is shorter. The cap exists because media does not only exhaust by throughput — carbon adsorbs slowly from standing water, biofilm develops on wet media regardless of flow, and a cartridge sitting half-used for three years is not half-fresh. A capacity calculation can only ever move a replacement earlier than the calendar, never later.
| Example | Rated capacity | Assumed use | Capacity says | What actually governs |
|---|---|---|---|---|
| Pitcher filter | 200 gal | 2 gal/day, couple | 100 days ≈ 3.3 months | Capacity. It lands close to the stated 2–4 months, so the two agree |
| Refrigerator filter | 300 gal | 3 gal/day incl. ice | 100 days ≈ 3.3 months | Capacity, well before the 6-month calendar. A busy ice-maker halves the label interval |
| Under-sink carbon | 10,000 gal | 4 gal/day drinking & cooking | 2,500 days ≈ 6.8 years | The calendar, by a mile. The 12-month cap governs and the rating is nearly irrelevant |
| Whole-house cartridge | 100,000 gal | 300 gal/day, whole home | 333 days ≈ 11 months | Capacity and calendar nearly coincide — but sediment can end it far earlier |
Assumptions: daily volumes are illustrative household figures, not measurements from your home — the under-sink and pitcher figures count only drinking and cooking water, while the whole-house figure counts everything including showers and laundry. Read the pattern rather than the numbers: on point-of-use filters the capacity rating usually binds; on large under-sink blocks the calendar binds; on whole-house cartridges sediment binds before either.
The relationship is a simple inverse: double the household’s use and you halve the service life. ⚠ The amber rows are the trap — two of the three exceed a twelve-month manufacturer cap, so the calculation must be capped rather than followed. Capacity can bring a replacement forward; it can never push one back.
Time, capacity, pressure drop or test result — which trigger wins?
Decision flow. Work top to bottom and stop at the first yes. Most owners run this list backwards — they wait for a taste change, which is the least reliable signal of the five.
[Replace now checklist]
- Meaningful flow reduction at the filtered outlet, not a vague sense that it is slower.
- A measurable rise in pressure drop across a whole-house housing. Two gauges either side of the filter turn this from a feeling into a number.
- Chlorine taste or smell returning — a late signal, but a definite one when it appears.
- Visible sediment loading on a cartridge you can see, or in a spin-down bowl.
- The filter-change indicator has turned. EPA’s guidance uses the traffic-light convention: green means working, yellow means change soon, red means change now Regulatory guidance.
- Rated capacity reached by the calculation above, even if nothing seems wrong.
- A failed water test — bacteria, lead, nitrate or any contaminant breakthrough.
- Physical damage, bypass, channelling or leakage. A cracked housing or a cartridge that water is routing around is not filtering at all, and it will look and taste completely normal.
Not every contaminant announces itself. Lead, nitrate, arsenic and PFAS are all tasteless, odourless and invisible at the concentrations that matter. Carbon typically loses its chlorine-taste capacity last — after it has stopped reducing the health-related contaminants you bought it for. Water that tastes good is evidence about the water’s taste and nothing else Editorial.
When does reduced flow not mean exhaustion?
Slow flow is the most common reason people buy a cartridge, and often the wrong one. Before spending anything: check the installation — a cartridge fitted the wrong way round or an O-ring pinched on reassembly restricts flow immediately. Inspect the prefilter, because on a multi-stage system the sediment stage usually loads first and blinds everything downstream. Confirm the valve position, including any bypass left half-closed after the last service. Measure the pressure drop rather than guessing at it. Read the manual for that model’s stated interval and any flush procedure. Test the water if the concern is health rather than flow. And call the utility or manufacturer — a pressure complaint that started the same week across your street is a mains problem, not a filter problem, and replacing cartridges will not fix it. Whole-house pressure loss also frequently traces to a failing pressure regulator or a clogged aerator, neither of which is a filter.
A manufacturer’s own replacement walkthrough for a multi-stage system. Worth watching before your first change — the recurring mistakes are over-tightening housings, forgetting to depressurise, and reusing a tired O-ring, and all three show up as leaks or flow loss that get blamed on the new cartridge.
How should you document maintenance?
Write the date on the cartridge in permanent marker as you fit it. It survives moves, sales and forgetfulness better than any app, and it answers the only question that matters when you are standing in front of the housing eight months later.
Beyond that, keep a single note per system recording the model and its rated capacity, the install date of each stage, the pressure reading when everything was new, and the date and result of your last water test. That baseline pressure reading is the one people skip and later wish they had — without it, “the pressure has dropped” is an impression rather than evidence.
Stop running your house on a generic six-month rule
Enter your filter types, installation dates, rated capacities and household usage to create a maintenance calendar based on your system — not a generic six-month rule. The calculator above handles a single component; the calendar tracks every stage you own and tells you which is due next.
When should you test instead of replacing?
Whenever the question is about safety rather than service life. A new cartridge answers “is this filter tired?” — it cannot answer “is my water contaminated?”, and swapping one to settle that question buys a false negative. After a boil-water advisory, after work on the well, after any change you cannot explain, and on any well annually, the correct next step is a laboratory test. Testing costs less than most people assume — often $10 to $30 per analyte at a state or county laboratory — and it is the only way to learn something a replacement cannot tell you.