How accurate is the label in your clothes?

The short version

• Two large studies put label inaccuracy between roughly a third and two-fifths: 37% mislabeled in the EU's 2026 market-surveillance campaign, and 41% likely to mislead in a 2019 Dutch government-commissioned scan of more than 10,000 discarded garments.
• Labels drift through substitution, blend drift, carried-forward paperwork, and rules that are required but rarely tested.
• Blends are the weak point, cotton-polyester most of all. Single-fiber claims carry the best odds available.
• Buying online raised the failure rate by about ten points against buying in a store.
• Thirty seconds at the rack improves your odds. Only a laboratory settles the question.

How often is a clothing label accurate?

Somewhere between roughly a third and two-fifths of the labels in those two samples did not match the fabric. That is the honest headline, and it comes with caveats worth stating.

The EU figure is the current one. Inspectors in eight countries tested 132 garments in 2026 against the composition printed on the tag, using laboratory methods rather than judgment, and 37% failed. The sample is small against the size of the market, which is part of the point: coordinated checks like this are rare enough that a single campaign becomes the reference number for years.

The Dutch figure is older and scoped differently. Those 10,000-plus garments were discarded, so the sample skews toward clothes made years before anything on a rack today. It is also far larger, because near-infrared scanning can cover volume that destructive lab testing cannot.

Neither study is a census. Both are samples, and both should be read as strong signals rather than a national statistic. What makes them worth taking seriously together is that they disagree by a few points rather than by an order of magnitude, despite testing different clothes, in different countries, with different instruments, years apart.

The full comparison, including what each method can and cannot see, is in "The 41% problem".

Why labels drift out of true

Most mislabeled garments are not the product of a decision anyone made in a room. They come out of a chain where checking is nobody's job.

• Substitution up the supply chain. A yarn runs short, a mill swaps in whatever is available or cheaper, and the cloth changes while the paperwork does not.
• Blend drift. Percentages on a blend label often describe an intended recipe rather than a measurement of the cloth that came off the loom. Small differences across several yarns compound into a number that was never quite true.
• Paperwork carried forward. Composition gets copied from a previous season's specification, or from a supplier declaration nobody retested. The claim outlives the garment it was written for.
• Weak enforcement. Fiber content is legally required. FTC textile rules in the US and the EU Textile Regulation both demand fiber names and percentages by weight. Neither regime tests routinely. Spot checks like the EU campaign are rare, and 132 garments is a rounding error against what enters the market in a week.

Honest error runs in both directions, and some of it does. In the Dutch data, 11% of labels understated the expensive fiber, meaning the garment was better than the tag claimed. But about a third of cotton-rich blends overstated their cotton content, and cotton is the fiber shoppers pay a premium for. When overstatement runs at roughly three times the rate of understatement, the gap between those two figures is the part that random error does not explain.

Where labels fail hardest

Accuracy is not spread evenly across a wardrobe. It tracks how complicated the claim is.

• Blends of natural and man-made fibers failed 64% of the time in the EU campaign. The more numbers a label carries, the more places one of them has to be off.
• Cotton-polyester was accurate only about 23% of the time in the Dutch scan of discarded garments. Fewer than one in four.
• Claims of 100% of a single natural fiber failed 15% of the time in the EU campaign. Those are the best odds on offer, and still not a guarantee.
• Baby clothing failed at 25% on a 36-item sample in the same campaign. A small sample, worth holding loosely, though it points the same direction as everything larger.

One number is easier to check and harder to shade than four. That is the whole mechanism, and it is why the composition line is worth reading as a risk score as much as a spec. A blend is not automatically worse fabric. It is a claim with worse odds of being accurate.

More on that pattern in "Blends are where labels go wrong".

Where you buy changes the odds

The EU campaign split its results by channel. Items bought online failed at 46%. Items bought in physical stores failed at 36%. Ten points is a real difference, and it comes from what each setting lets you do.

In a store you can hold the fabric, find the sewn-in tag, and see whether the swing ticket and the care label agree with each other. Online, you have a text field typed by whoever built the product page, and sometimes not even that.

Shoppers already sense the gap. Cotton Incorporated's Lifestyle Monitor finds 63% saying fiber content matters to their purchase decision while only 42% always or usually check the label, and 52% say knowing fiber content online would influence what they buy. A product page that omits fiber content altogether has made a choice about that, and the silence is worth reading as a flag of its own.

The channel gap gets a practical walkthrough in "How to read a fabric label in 30 seconds".

What you can check yourself in 30 seconds

You cannot measure fiber content at a rack. You can improve your odds, and it takes about half a minute.

• Find the fiber line, not the care symbols. It is legally required in both the US and the EU, and listed in descending order by weight, so the first fiber named is the one there is most of.
• Count the fibers. One is better odds than three. A top sold as cotton that reads "52% cotton, 48% polyester" is a different garment from one that reads "95% cotton, 5% elastane", even though both get marketed the same way.
• Cross-check the statements. The sewn-in tag, the swing ticket and the product page should all say the same thing. Disagreement between any two of them tells you the composition was not carefully controlled.
• Note the red flags. A strong chemical smell out of the packaging. Ultra-saturated color on a very cheap item. Claims like "wrinkle-free", "stain-proof" or "antimicrobial", which describe a chemical finish rather than a weave.
• Look for a certificate number, not a logo. OEKO-TEX® Standard 100 and GOTS® both issue numbers anyone can verify free in their own public databases. A reassuring green logo with no number beside it is unverifiable.

None of this settles composition. It sorts garments into "plausible" and "unchecked", which is already more quality control than most shopping trips manage.

The long version is in "How to read a fabric label in 30 seconds", and the certification test has a piece of its own: "A logo without a number means nothing".

What only a lab can settle

Fiber composition is a measurement, and measurements come from instruments. The methods are old, standardized, and boring in the way good methods are.

Identification comes first. Under methods like AATCC 20, an analyst establishes which fibers are present using microscopy, burn behavior and solubility. Cotton is a twisted ribbon under magnification; polyester is a smooth rod.

The percentages come from quantitative analysis, standardized in the ISO 1833 series and AATCC 20A. A solvent is chosen that dissolves one fiber and leaves the other intact. The sample is weighed, dissolved, dried, and weighed again, and the difference is the dissolved fiber's share. A scale, a solvent, and arithmetic, which is why a credible report cites the exact method number it used.

Who runs the test matters as much as which test it is. ISO 17025 is the accreditation standard for testing laboratories: it means an independent body has assessed the lab's methods, equipment and competence rather than taking its word for them. Lynen does not operate a laboratory. We commission independent accredited laboratories and publish what they send back, including when the result does not match the label.

Two limits belong in plain sight. The test is destructive, so labs analyze representative samples of a product rather than the specific unit hanging in your wardrobe. A result is strong evidence about that product, not a certificate for your individual garment. And each verification costs real money, so coverage grows garment by garment rather than arriving complete.

The bench work is described step by step in "What happens in the lab".

What is coming, and what it will not fix

Two things are moving.

The EU's digital product passport is expected to reach textiles around 2027, giving garments a scannable record of what they are made of, where they came from, and how to care for them. The format is a genuine step up. The source is unchanged: a brand-declared record, the same claim better delivered. Moving a claim from a stitched-in tag to a QR code does not make it accurate, which is why independent checking gets more useful in a passport world rather than less.

Durability is the other. Pilling and color fading are the top two measured causes of garment failure, according to a UK WRAP and Defra survey analyzed in the Journal of Cleaner Production in 2022, and both are largely written into a textile before it is sold. Wear-test results are not live in Lynen. Wear testing is coming to verified garments, and we will say so when it arrives rather than before.

When we have not checked a garment, it shows as not yet verified, not as a pass.