Honest limits
Are microplastic test kits accurate?
For the question they're designed to answer — how many plastic particles ≥1 µm are in this 100 mL sample — yes, within ±15% repeatability. For the questions they aren't designed to answer — which polymer, how many nanoplastics, is it safe to drink — no, and we say so on every page. Here's the honest accuracy picture, what the published validation looks like, and where our microplastic test kit sits relative to a gold-standard lab.
What “accurate” actually means here
Accuracy isn't one number — it's different per question. A microplastic count from any method has at least three independent dimensions of accuracy: sensitivity (the smallest particle it can see), specificity (the rate of false positives), and repeatability (how close two runs of the same sample come).
The Nile Red staining method — published in Scientific Reports (Maes et al., 2017), validated at scale in Frontiers in Chemistry (Mason et al., 2018), and used in dozens of follow-on surveys — has documented specificity (Nile Red binds plastic and a handful of other hydrophobic particles, not minerals or salts) and documented repeatability (±10–15% inter-run on the same sample). Sensitivity depends on filter pore size; at 1 µm the kit catches the microplastic fraction.
An in-home kit is more accurate than the marketing on most filter packaging (which makes broad “removes contaminants” claims without specifying which) and less accurate than a lab using gold-standard instruments like py-GC/MS or micro-FTIR (which identify polymer chemistry the kit can't). Use the right tool for the question.
Accuracy by question
| Question | In-home kit | Lab (gold standard) |
|---|---|---|
| Particle count (≥1 µm) How much plastic is in this sample? | Reliable. Matches lab Nile Red protocol. | Reliable. Same method, instrument-imaged. |
| Polymer identification Is this PE or PET or PVC? | Not supported. Requires py-GC/MS. | Available on premium tiers. Required: py-GC/MS or micro-FTIR. |
| Nanoplastic detection (<1 µm) How many sub-micron particles? | Not supported. 1 µm filter floor. | Available with stimulated Raman scattering / EM. Specialized labs only. |
| Repeatability (same sample, two runs) Is the count stable? | ±15% inter-run variability | ±5–10% inter-run variability |
| Side-by-side comparison (sample A vs B) Tap vs filter, bottled vs tap, etc. | Excellent — same kit, same conditions, same observer. | Possible but expensive per sample. |
| Regulatory / legal reporting Will a court or utility accept the result? | Not appropriate. No chain of custody, no accreditation. | Yes (accredited labs only). |
What the kit catches reliably
Fragments and films, 1 µm to ~5 mm. The bread-and-butter of microplastic surveys. Bottle wall fragments, pipe-shed flakes, packaging films.
Fibers ≥1 µm in diameter. Synthetic textile fibers (polyester, nylon, acrylic) — the dominant microplastic class in indoor air and tap water per multiple published surveys.
Pellets and beads. Personal care products historically contained PE microbeads; they were banned in US rinse-off products in 2015 but persist in older formulations and in some non-rinse products.
What the kit doesn't catch
Nanoplastics below 1 µm. The 1 µm filter pore is the floor, and no optical method gets under it. This is the size class that matters most for the body: under 1 µm, particles cross cell membranes, the gut lining, and the blood-brain barrier. The 2024 PNAS paper from Qian et al. used stimulated Raman scattering to count down to ~100 nm in bottled water and reported ~240,000 particles per liter, ~90% of them nanoplastics — ~1000x larger than micron-scale surveys. Those particles are real but invisible to any consumer kit, including ours; the micron count is the visible proxy for them. See the nanoplastics explainer for why size changes the health picture.
Polymer chemistry. Telling PE from PET from PP from PVC requires py-GC/MS, micro-FTIR, or Raman spectroscopy. These are lab instruments costing $25k+. The kit confirms presence and quantity; the lab adds identity.
Particle shape distribution. Whether your count is mostly fragments vs fibers vs films requires manual microscope inspection or AI-based image classification. Some hints come from the photo (long thin streaks = fibers, round dots = fragments) but it's not a full taxonomy.
The error source every method shares: contamination
Airborne synthetic fibers settle into open containers — in your home and in research labs, which is why published studies run procedural blanks and work in clean rooms. An in-home kit can't do either, so the protocol leans on two defenses: keep the container covered during waits, and run the two included tests side-by-side so ambient contamination cancels out of the comparison. A single-digit count is within the range contamination alone can explain; large counts and large differences between samples are the trustworthy signals. The complete accounting of every known error source is on the limitations page.
We've also documented our own procedural floor: clean, filtered water run through the full workflow returns 0–3 particles on the filter. Counts meaningfully above that floor are real load from the sample, not an artifact of the procedure.
Filter membrane consistency
The honest edge of an in-home test is the filter membrane itself. Membranes come from different vendors and their quality varies batch to batch — some carry a little trace material of their own, and it's hard to QA an individual membrane before it ships. Across the 250+ tests we've run since inception we've learned which suppliers produce the cleanest membranes, and we're continually working to move to higher-quality filters with lower background. Because a membrane can carry a small baseline of its own, read a single count as an upper bound, not an exact figure.
How to sharpen your count. You can account for the membrane's baseline yourself with a quick blank: before you run your sample, look at an unused filter under the digital microscope and do a rough count of any glowing dots already on it, then run your water, count again, and subtract the first number. A formal procedural blank is the lab version of the same idea — the quick visual check gets you most of the way there. Running the two included tests side-by-side does this for you too: the membrane baseline hits both equally, so the difference between them is the trustworthy signal.
The validation work we're doing
The underlying method is peer-reviewed. The packaged kit is a faithful implementation of the same fluorescent-staining approach the published protocols validated. But that's an argument by construction; an independent comparison against a gold-standard instrument on identical samples is stronger evidence.
That comparison is in progress. When the validation data lands, it goes on the methodology page with the raw side-by-side counts. Until then, we point at the published literature and at our own internal reproducibility tests.
FAQ
How accurate is an in-home microplastic test kit compared to a lab?
At the ~1 µm-and-up size class, an in-home Nile Red kit and a lab using the same Nile Red protocol return comparable particle counts — that's the whole point of standardizing the method (Maes 2017, Mason 2018). What a lab adds is polymer identification (using py-GC/MS or micro-FTIR) and nanoplastic detection (sub-1 µm). For "how many microplastic particles are in this water," the answers converge. For "what polymer" or "how many nanoplastics," only the lab can answer.
Does Nile Red give false positives?
It can, on samples rich in organic matter — tannins from tea, lipids from formula, biofilm from old plumbing. Organic matter is partially hydrophobic and binds some Nile Red. The standard lab control is a hydrogen peroxide digestion step that oxidizes those organics into water-soluble fragments that pass through the filter, but the current kit does not include it — digestion-strength peroxide is a hazardous material to ship, so for now we ship a rapid, safe screening test without it and are doing R&D on lowering the false-positive rate another way. So treat counts on organic-rich samples as upper bounds. Either way, Nile Red is a screening method — a count is suspected microplastics, not chemically confirmed ones. The full error budget is documented at thewatertest.com/limitations.
How consistent are the filter membranes?
This is the honest edge of an in-home test. Membrane filters come from different vendors and their quality varies batch to batch — some carry a little trace material of their own, and it's hard to QA an individual membrane before it ships. Across the 250+ tests we've run since inception we've learned which suppliers produce the cleanest membranes, and we're continually working to move to higher-quality filters with lower background. Because a membrane can carry a small baseline, treat any single count as an upper bound rather than an exact figure. You can account for it yourself with a quick blank: look at an unused filter under the digital microscope first, do a rough count of any glowing dots, then subtract that from your post-dye count. Running the two included tests side-by-side does this for you too — the membrane baseline hits both equally, so the difference is the trustworthy signal.
Does the kit miss small plastic particles?
Yes, anything below the filter's 1 µm pore size passes through and isn't counted. This is the nanoplastic range. The 2024 Columbia/Rutgers PNAS paper used stimulated Raman scattering to count nanoplastics in bottled water and reported ~240,000 particles per liter, ~90% of which were sub-1 µm. None of that would be caught by an in-home kit (or by most published Nile Red surveys). We say so openly — the kit measures the microplastic fraction, not the nanoplastic fraction.
How does my count compare to published numbers?
The 2018 Orb Media bottled-water survey (Mason et al.) averaged ~325 particles per liter, which converts arithmetically to about 33 per 100 mL (the volume the kit tests): ~10/L above 100 µm (~1 per 100 mL) and ~315/L in the 6.5–100 µm range (~31.5 per 100 mL). Treat that ~33 as a theoretical reference point, not a prediction — it's a straight division of a study average, and a real result varies widely by brand, bottle age, and handling. Note the study's imaging reliably counted only down to ~6.5 µm even though its filters caught finer debris, so 325/L is an above-that-size figure and the real number is higher. Our kit counts particles ≥1 µm — finer than that floor. As rough reference points on the kit: a clean distilled-water sample returns <5 particles; a typical tap returns 10–100; a typical bottled water returns 50–500; a year-old bottled water in a hot environment can exceed 1,000. Below all of this sits the nanoplastic range no optical kit reaches — the 2024 Columbia/Rutgers PNAS study counted ~240,000 particles per liter down to ~100 nm, ~90% of them sub-1 µm.
Has the kit been validated against a lab?
The underlying method has been validated extensively in the peer-reviewed literature (Maes 2017, Mason 2018, plus dozens of follow-on studies). We are working on an independent third-party validation comparing kit counts to py-GC/MS counts on identical samples; results will be published on the methodology page when complete. Until then, the kit's accuracy claim is that it implements the published Nile Red protocol faithfully — same dye, same filter, same wavelength, same imaging — at consumer scale.
Will two people running the same sample get the same count?
Within ±10–20%, yes — particle counting always has inter-observer variability because the human eye decides whether a faint dot crosses threshold. Our upload pipeline runs an automated count on your photo using the same threshold logic for everyone, which removes that variability. Run the test on the same sample three times and you'll get three counts within ~15% of each other (this is the published reproducibility range for Nile Red on filter membranes).