How we score microplastic exposure

Every product in our database carries two numbers rather than one. That is the whole methodology in a sentence, and it is the thing that makes these pages different from a scanner app that shows you a red circle and a warning triangle.

Number one

Exposure, 0–100

How much plastic contact this product involves — what material touches the contents, at what temperature, for how long, under what mechanical stress. Higher means more contact.

Number two

Confidence, 0–100%

How much anyone has actually measured about this specific product. A product nobody has tested can be scored — it cannot be scored confidently, and we will not pretend otherwise.

What the exposure score is not. It is not a health verdict, a safety rating, or a prediction of harm. It measures plastic contact, which is a physical property of a product. Whether a given level of exposure matters to human health is an open scientific question — the FDA's current position is that the evidence does not demonstrate a risk at the levels found in food, while research such as the 2024 NEJM plaque study gives real reasons to keep looking. Both of those are true at once, and any site telling you only one of them is selling something.

The model

Five components, adding to 100

The score is additive and the components are rendered on every product page. Nothing is stored as a total — the page adds the parts each time it builds, so a page cannot drift from the reasoning behind it, and you can always see which component is carrying the number.

Packaging contact

0–30

What material touches the contents, over how much surface area, for how long. Glass and steel score near zero here; a film pressed against fatty food scores near the top.

Use conditions

0–25

Heat, abrasion, flexing, repeated opening, reuse beyond design life. The same bottle scores differently in a fridge and in a hot car.

Measured release

0–25

What a laboratory actually counted, for this product or its category. A high number here means real particles were measured, not that the material looks suspicious.

Polymer behaviour

0–10

How this specific polymer behaves — brittleness, fat solubility of its additives, thermal stability, whether it is a polymer named in existing restrictions.

Processing

0–10

What happened before you bought it. Retort sterilisation inside the pack, hot-filling, and pressurised brewing all happen at the most migration-favourable moment.

Reading the number

What each band means

Exposure score bands and what each one means
Band	Score	What to do with it
Low	0–24	Little or no plastic in contact with the contents. This is the option to switch toward.
Moderate	25–49	Some plastic contact, but limited by material, temperature or contact time. Reasonable as a default.
Elevated	50–74	Meaningful plastic contact under conditions known to increase release. Worth swapping if a clean alternative is easy.
High	75–100	Heat, abrasion or direct contact with fatty food, usually more than one at once. The highest-value swaps live here.

Evidence grades

A, B, C, D — and what each one is admitting

The grade sets a hard ceiling on the confidence a page may claim. It is enforced in code rather than by editorial discipline: a grade D page cannot report 90% confidence even if someone types it into the data. Across the 132 products currently in the database, the distribution is 12 grade A, 29 grade B, 70 grade C and 21 grade D.

Grade A — direct measurement

confidence 85–95%

This exact product, or an item identical to it, has been measured for particle release in a peer-reviewed study. The number is anchored to a laboratory result rather than to a material.

Grade B — category measurement

confidence 65–80%

The product category and packaging format have been measured, but this specific brand or SKU has not. We apply the category result and say so. This is the most common grade, and honestly so.

Grade C — polymer evidence

confidence 50–64%

No measurement of the product or its category exists. The score rests on published evidence about how this polymer behaves under these conditions — heat, abrasion, contact time, fat content.

Grade D — packaging inference

confidence 35–49%

The score is inferred from the packaging and how the product is used, with no measurement at any level. Directionally useful for choosing between two options; not a number to quote as fact.

The part that looks like a bug

An unmeasured product scores lower, on purpose

One of the five components is measured release, and it only scores points where a laboratory actually counted particles. So two identical PET bottles — one in a tested cohort, one not — will not get the same number. The tested one scores higher.

That is deliberate. We will not award exposure points for particles nobody has counted, because doing so would let us assign any number we liked to anything and call it data. The gap is carried by the confidence figure instead: a grade B page saying "60, at 72% confidence" is telling you the number can move, and in which direction. Comparisons across evidence grades should be read with that in mind, which is why every head-to-head page shows both grades side by side.

Independence

What money can and cannot buy here

No brand pays for a score. There is no product, tier or arrangement that changes a number in this database. Scores are computed from the components above and nothing else touches them.
No sponsored placements in rankings. Category orderings are sorted by the computed score. There is no promoted slot and no paid inclusion.
Affiliate links never influence a score. Where we earn a commission we say so on the page and in our affiliate disclosure. A product does not move up for being commissionable, and several of our lowest-scoring recommendations earn us nothing at all.
Brands can dispute a score, not buy one. We publish the components, so a dispute has something specific to argue with. Send evidence and we will re-score in public, including when that means correcting ourselves.

Full detail: affiliate disclosure · how the brand certification programme is kept separate

Corrections

We publish our own mistakes

A corrections policy is worth nothing without corrections in it. These are ours, newest first.

Correction · September 23, 2026

“No plastic-free drip machine” was out of date

Our plastic-free coffee maker guide said no automatic drip machine has a plastic-free option. Simply Good Coffee has sold a drip brewer with a glass tank and a stainless water path since April 2025, and Caraway announced one on September 22, 2026. The guide now covers both, with what each maker does and doesn't disclose (Simply Good's includes silicone seals; Caraway hasn't said what its seals are). It also said the OXO Brew has a plastic tank and basket, which OXO doesn't publish; that claim was removed.

Correction · September 23, 2026

Coffee pod rankings presented as measured

Our coffee pod guide said plastic K-Cups release the most microplastics “in published testing” and that Nespresso pods release fewer “in published comparisons”. No study has compared pod types particle for particle, and the one count of capsule coffee did not isolate the pod from the machine. The guide now ranks pods by how much plastic sits in the hot-water path, and says so. It was merged with our separate K-Cup article, which now redirects to it.

Correction · September 23, 2026

The silicone air-fryer mould study, overstated

Our non-toxic air fryer guide said a September 2026 study found silicone air-fryer moulds releasing a phthalate and siloxanes over the limit “in a fatty-food simulant”. The exceedances were in the study's harshest test, a dry-food simulant (Tenax) at 175 °C; in water-based simulants the phthalate stayed below the EU limit, and by the third use levels were below all the criteria the authors applied. The guide also said disposable parchment liners are mostly silicone-coated; some brands use wax and some don't disclose a coating. Both were corrected.

Correction · September 23, 2026

Glass-bottled water does not test near-zero

Our bottled water ranking said glass-bottled brands such as Voss, Mountain Valley and Ferrarelle “consistently test near-zero” for microplastics and that glass-bottled water “releases zero microplastics”. None of those brands has a published microplastic test, and the three studies that compared glass and plastic bottles directly (ANSES 2025, Schymanski et al. 2018, Oßmann et al. 2018) found glass-bottled water with as many particles as PET, or more, traced to the caps. The glass itself is inert; the closure is not. The ranking also gave Voss's source as Norway only, while Voss's US quality report lists US East and US West sources. The ranking was corrected and the evidence is in our Voss article.