We Scored 132 Products for Microplastics. Only 12 Have Ever Been Measured.

Every product page on this site carries a number from 0 to 100, an evidence grade from A to D, and a confidence percentage. The grade is the part almost nobody reads, and it is the part that decides what the number is worth. So we audited our own database and counted.

What 132 products actually rest on

Grade A means a peer-reviewed measurement exists for that exact product or an item identical to it. Grade D means we are reasoning from the polymer and the way the thing is used, with no direct study of anything close to it.

A majority of the database — 53% — sits at grade C. If you have ever looked up a product here and taken the number at face value, the honest framing is that you were most likely reading a well-argued estimate rather than a laboratory result. That is what the grade has always been there to tell you. We just had not added it up before.

The measured 9% is not evenly spread, either. Five of the twelve grade-A products are bottled waters, which is a direct consequence of the 2018 Mason study and the 2024 PNAS work having concentrated on that one category. Personal care, coffee, food storage, drinkware, household and water filters have no grade-A product at all between them — 56 products, not one of them measured directly.

The finding we did not expect

Measured products score worse. Grade-A items average 62.3 out of 100. Grade-C items average 39.9. That is a 22-point gap, and it points the opposite way from the comfortable assumption that testing mostly reassures.

The obvious objection is that this is circular. One of the five scoring components is measured release, worth up to 25 points, and by construction it can only score high when a measurement exists. So we removed it and recalculated:

The gap narrows but does not close. Strip the measurement out entirely and grade-A products still score 41.6 against 32 to 35 for everything else. It shows up in components that are scored independently of whether a study exists: packaging contact averages 18.6 for grade-A items versus 14.0 to 15.6 elsewhere, and use conditions 15.9 versus 11.8 to 13.8.

Why that is selection bias, not a discovery

The tempting read is “measurement reveals exposure that reasoning misses.” We do not think that is what this shows, and it would be a convenient conclusion for a company that sells a scanner, so it is worth being careful.

The likelier explanation is that researchers measure what they already suspect is worst. Nobody funds a study of a wooden cutting board. They fund studies of pyramid tea bags steeped at 95°C, of polypropylene baby bottles under shaking and heat, of bottled water that has sat in a warehouse. Those are already high-contact, high-temperature, high-agitation items before anyone measures them — which is exactly why they score high on packaging and use conditions independently of the lab result.

The practical consequence matters more than the statistic. Our grade-C numbers are not simply blurrier versions of our grade-A numbers. They describe a systematically different population of products — one that skews toward lower-contact, lower-temperature items that nobody has bothered to test. Treating the two sets as interchangeable, in either direction, would be a mistake.

What scores worst, by category

Averaged across every product we have scored in each category, with the count of directly measured items alongside:

Two things stand out. Textiles top the table on a mean of 58.3 from only seven products, which is the smallest sample here and should be read as a signal to test more rather than a settled finding. And bottled water, the most-studied category on the list by a distance, sits mid-table at 42.7 — below packaged food and below baby products, neither of which has anything like the same evidence behind it.

The extremes are worth naming because they are where the reasoning is easiest to check. The highest score in the database is a pyramid mesh tea bag at 87: woven plastic, submerged, at close to boiling point, with nothing between the polymer and the drink. The lowest is a wooden cutting board at 8, followed by loose leaf tea with a steel infuser at 9 and a stainless lunch box at 10. In every one of those cases the score is low for the same reason — there is no polymer in the contact path at all.

What we are doing about it

The obvious response to “most of your numbers are reasoned” is to publish fewer numbers. We think that is the wrong call. A reasoned estimate with its reasoning shown, its confidence stated and its grade on the page is more useful than silence, provided nobody is misled about which kind of number they are reading.

What it does change is where the next work goes. The six categories with zero measured products are where our numbers are weakest, and three of them — personal care, coffee and drinkware — are among the most-scanned categories in the app. That is the gap worth closing, and it is a research problem rather than a writing one. You can read the full scoring method, component by component, on our methodology page.