Everything this test can't do.

The kit uses Nile Red fluorescence staining. Nile Red binds to hydrophobic (water-repelling) surfaces and fluoresces when it does. Common plastics — PE, PP, PS, PET, nylon — are hydrophobic, so they stain and glow. But the dye reports hydrophobicity, not polymer chemistry. The foundational paper for this method (Maes et al., 2017, Scientific Reports) describes it, in its own title, as a “rapid-screening approach.” That is exactly what it is, and exactly how we use it.

A glowing particle on your filter is a suspected microplastic: a hydrophobic particle in the microplastic size range that stained and glowed. In drinking water that population is dominated by plastic. It is not chemically confirmed plastic. Confirmation requires micro-FTIR, Raman spectroscopy, or pyrolysis GC/MS — five-figure laboratory instruments.

Because Nile Red stains hydrophobicity generally, non-plastic hydrophobic material can fluoresce too. The documented offenders in the literature: lipids and fats, natural waxes, rubber fragments, some adhesives, and certain natural organic matter (biofilm, algal debris).

The standard laboratory control for this is a hydrogen peroxide digestion step, which oxidizes organic material into fragments that pass through the filter unstained. The current kit does not include this step. Digestion-strength peroxide is regulated as a hazardous material for shipping, and we've chosen — for now — to ship a rapid, safe at-home screening test without it rather than ship hazmat to consumers. We're actively doing R&D on ways to lower the false-positive rate without a hazardous reagent.

The practical consequence: without a digestion step, organic-rich samples (tea, formula, pond water, water from old biofilm-lined plumbing) will read high, because non-plastic hydrophobic material that a digest would have removed stays on the filter and stains. Treat those counts as upper bounds, not measurements.

What we can't yet give you is a kit-specific false positive rate. That requires running known-clean and known-spiked samples through the kit at scale — the spike-recovery study on our validation roadmap — and publishing the recovery curve. Until that lands, treat every count as an estimate with real, unquantified uncertainty, not a measurement with an error bar.

This is the limitation that matters most, and the one microplastics researchers flag first. Indoor environments are full of microplastics — chiefly synthetic textile fibers that settle out of the air onto every surface, including the inside of a jar that's been sitting on a shelf and the surface of your sample while it sits open during the protocol's wait steps. Those particles stain and glow exactly like particles that came out of your tap, because they are the same thing. Published laboratory studies fight this with clean rooms, cotton lab coats, and glassware muffled at high temperature — none of which exist in a kitchen, bathroom, or garage.

What the protocol does about it: you rinse the container three times with the same water you're about to test (washing out settled material without introducing anything the test water doesn't already contain), keep the container covered during waits, and push the sample through the filter promptly rather than letting it sit open.

What the protocol can't do: reduce airborne deposition to zero, or tell you after the fact which particles came from the water and which came from the room. This is why we tell you not to over-read low counts, and why the kit is built around comparison — more on both below.

In laboratory QA/QC, a procedural blank is clean, filtered laboratory water carried through the entire sampling and processing workflow — same container, same equipment, same air exposure, same handling as the real sample. Its purpose is to quantify contamination introduced by the procedure itself. Whatever the blank counts, you know your method added.

This kit does not include one, and we won't dress anything else up in that language. Running a true procedural blank at home would mean shipping every customer sealed laboratory-grade water and asking them to spend a test on it — roughly doubling the cost of a kit whose reason to exist is affordability, to answer a question most users aren't asking. Our customers are homeowners and filter installers, not lab technicians, and the protocol is built for them. What we have done is run procedural blanks on our end: clean, filtered water carried through the full protocol consistently returns 0–3 particles on the filter. That is the floor the procedure itself can introduce — counts materially above it are real sample load.

The practical consequence, stated plainly: there is no per-run number separating what came from the water from what the procedure added. Contamination is controlled by protocol (the rinse, the covered waits, prompt filtration) but not quantified by a blank. That is a real limitation — and one shared by every consumer test in this category. The consumer-appropriate control is the side-by-side comparison design described in section 9: two samples run identically, where procedure-introduced contamination cancels out of the difference.

The filter membrane retains particles down to about 1 µm. The imaging — a phone camera photographing the filter under blue light — reliably resolves particles of roughly 10 µm and up. Particles between those bounds are captured on the filter but may be under-counted in the photo. Everything below 1 µm passes through the filter entirely.

That sub-micron fraction is the nanoplastic range, and by count it is most of what's in water: the 2024 PNAS study (Qian et al.) counted ~240,000 particles per liter in bottled water with ~90% of them below 1 µm. No consumer test reaches that fraction — it takes stimulated Raman scattering or electron microscopy. Your count is a visible proxy for a much larger invisible population. See the nanoplastics explainer for why size changes the health picture.

PE, PET, PP, PS, and nylon all stain the same. The kit cannot tell you which polymer a particle is, which matters if you're trying to trace a source (bottle wall vs. textile fiber vs. pipe shed). Shape gives a hint — long thin streaks are usually fibers, round dots usually fragments — but that's a hint, not an identification.

Run the same water twice and the counts will differ — by roughly ±15% in our internal repeat runs, consistent with the published inter-run range for Nile Red on filter membranes. Sources: which particles land where on the filter, focus and exposure differences between photos, and threshold effects on faint particles. Our automated counter applies the same threshold logic to every photo, which removes person-to-person judgment but not photo-to-photo variance. This is also why the instructions are strict about Night mode — an underexposed photo undercounts, full stop.

The underlying method is peer-reviewed (Maes et al. 2017; Mason et al. 2018; Leonard et al. 2022). The kit is a faithful packaging of that method. The kit itself has not been independently validated by an accredited laboratory. Spike-recovery calibration is in progress; an independent split-sample comparison against an FTIR/Raman lab and a cross-kit reproducibility study are planned but not started. The current status of each is always on the methodology page.

Results carry no chain of custody and no accreditation. They are not appropriate for regulatory compliance, litigation, or medical decisions, and we decline requests to use them that way.

Trust differences more than absolutes. The kit ships two tests so you can run a comparison — tap vs. filtered, tap vs. bottled — same day, same room, same handling. Ambient contamination and method bias hit both samples roughly equally, so they largely cancel out of the difference. “My filter cut the count by 80%” is a far more defensible statement than “my water has exactly 47 particles.”

Don't over-read low counts. A handful of particles is within the range that contamination alone can explain. The meaningful signals are large counts, large differences between samples, and consistent results across repeat runs.

Treat the number as an estimate. Until the spike-recovery work publishes a real error bar, every count is a screening estimate of suspected microplastics — useful for comparison, education, and deciding whether to investigate further. Not a laboratory measurement.

If you work on microplastics and see something here that's wrong, understated, or missing — tell us. Several changes to this kit's protocol and this site's language came directly from researchers who took the time to write. Email hello@thewatertest.com. Every public result and its raw image is on thewatermap.com if you want to audit the data itself.