How Often Should You Change the Water?

Written 2026-10-01

For a 60cm standard tank (about 65L) with a mid-size canister filter and 10 neon tetras, changing 20–30% of the water every week kept nitrate under the safety line for all 90 days. Skipping water changes entirely let nitrate climb to about 57ppm (test-kit units) by day 90, crossing the roughly 49ppm safety line, and afterward the buffering capacity (KH) ran out and the pH crashed. Changing a big 50% once a month never crossed the line, but it climbed to about 43ppm right before each change — much closer to the danger line than the weekly habit.

A filter can’t remove nitrate

The bacteria living in a filter only oxidize ammonia into nitrite, then nitrite into nitrate. They don’t remove nitrate itself. The only ways to bring nitrate down are replacing the water (a water change), uptake by live plants, and a small amount of denitrification in the oxygen-free layer deep in the substrate. So no matter how good the filter is, nitrate keeps building up as long as fish are fed and produce waste, unless some water gets changed.

Three water-change habits, run for 90 days

Everything else was held constant — a 60cm standard tank, a mid-size canister filter, cycling finished by day 15 before adding 10 neon tetras, and no plants (to keep plant nitrate uptake out of the comparison). Only the water-change habit changed: ① none at all ② 25% every week starting day 15 ③ 50% once a month starting day 30. In all three, every fish survived the full 90 days — but how the water quality moved over time was completely different.

HabitPeak nitrate (test-kit ppm)Result at day 90
None at all57.1ppm — over the 49ppm lineGrade C · "pH crash — buffering lost"
25% every week26ppmGrade B · "Stable run"
50% once a month43.6ppmGrade B · "Stable run"

Small and frequent vs. big and rare — why aren’t they the same?

Weekly 25% and monthly 50% remove a similar total amount of water over a month. Yet their peak nitrate readings were quite different — 26ppm versus 43.6ppm. Nitrate keeps accumulating between water changes without pause, so the longer the gap, the higher it climbs right before the next change. The monthly habit was spending its peak moments only about 5ppm below the 49ppm safety line. Changing less water more often keeps the swings smaller, so even when the average is similar, the tank spends far less time near the danger line.

Skip water changes entirely, and something else arrives before nitrate does

The tank that never had a water change didn’t lose a single fish in 90 days, but that doesn’t mean it was fine. Nitrate crossed the safety line by day 70, and the KH (carbonate hardness) that tap water would normally top up kept getting consumed with nothing replacing it, until it ran out. That collapsed the tank’s buffering capacity and crashed the pH — and once that happened, nitrification itself slowed down, and ammonia, which had been close to zero for most of the run, started climbing again, reaching about 2.3ppm. That buffering collapse is its own complicated story worth a separate guide, so here we’ll just note the result: skipping water changes breaks more than just nitrate control.

So how much, and how often?

Based on this comparison, 20–30% every week is a safe default. Changing a little often keeps nitrate lower and its swings smaller than putting it off and changing a lot at once. That said, changing too much in one go has its own risks — adding a large volume without a dechlorinator damages the filter bacteria right away, and adding a lot of unmatched tap water pulls the temperature down with it, which can trigger disease in species sensitive to swings within 24 hours. In Water Kiwoogi you can set the water-change percentage, interval, first day, dechlorination and temperature matching separately and run the simulation, so it’s worth plugging in your own tank’s numbers and comparing.

The button below runs the 60cm standard tank · mid-size canister filter · 10 neon tetras · 25% weekly water change tank from this guide.

Run this tank yourself

The numbers in this guide come from the Water Kiwoogi simulator model; real tanks can differ.