The PFAS-Free Claim

The Hanger

On a rack in a shop on Regent Street, a pair of performance leggings hangs from a plastic clip. The swing ticket carries a short-form claim. The claim is one word, or two, or three, depending on the brand: PFAS-free. Created without PFAS. Non-toxic. The word does not attach to a lab. It does not name a method. It does not cite a detection limit, a sampling regime, or a jurisdiction. It sits on a piece of card, alone.

The corporate sustainability page should be the place where that word unpacks into a procedure. A sample drawn. A method run. A threshold cleared. An auditor's signature. The word on the ticket should be the compressed output of all of that.

On the Lululemon corporate page, under the heading Created without PFAS -- what to know about lululemon's products, the procedure appears. It is three sentences long. The three sentences sit next to each other on the published page, and they are doing more work than the consumer reading them will recognise at a glance -- because the consumer is reading in the shopping register, and the sentences were written in a different one.

The word on the hanger is not a result. It is a compression of a paragraph, underneath which is a rulebook, underneath which are five rulebooks, none of which the consumer has been given.

The Paragraph

Every Fortune 500 sustainability page has sentences in the same register. This is not a Lululemon specialty; it is a class of corporate-epistemic language documented across a generation of disclosures in Hyland's hedging typology, Fairclough's promotional-discourse analysis, and the OECD's work on ESG communication.¹ Lululemon is the illustration here because the Texas Attorney General's Civil Investigative Demand, dated 2026-04-13, makes the text a public-record artifact. The pattern is industry-wide.

The verbatim three-sentence stack on the corporate page, as accessed in April 2026:

"lululemon does not use PFAS in products today."

"Our ongoing focus is to help prevent the unintentional reintroduction of PFAS into our products through ongoing testing, monitoring, and collaboration with suppliers and third parties."

"We require all our vendors to regularly conduct testing for restricted substances, including PFAS, by credible third-party agencies to confirm ongoing compliance."²

Three sentences. Three grammatical registers, stacked in a specific order.

Sentence one: "does not use PFAS in products today." The verb is does not use -- active, agent-bounded, volitional. The adverb is today -- present-tense, moment-pinned. The claim is about what the company does with intent, at one point on a clock. It is silent on yesterday's inventory. It is silent on tomorrow's production run.

Sentence two: "Our ongoing focus is to help prevent the unintentional reintroduction of PFAS into our products through ongoing testing, monitoring, and collaboration with suppliers and third parties." The verb is to help prevent. The qualifier is unintentional. The sentence acknowledges, in its own language, that reintroduction is possible -- and frames any reintroduction as accidental. The trailing clause adds three named mechanisms: testing, monitoring, and collaboration. None of the three is direct testing by the company. Testing is delegated to unnamed parties; monitoring is unspecified in scope; collaboration is with suppliers and third parties -- the same intermediaries who produce the chemistry. The mechanisms are named. The agent performing them is not the company.

Sentence three: "We require all our vendors to regularly conduct testing ... by credible third-party agencies to confirm ongoing compliance." Testing is performed by our vendors -- not the company. The labs are credible third-party agencies -- unnamed. The purpose is to confirm compliance -- not to test for presence. Three intermediaries stand between the sentence on the page and any test tube anywhere.

The claim is not written to be false. It is written in a grammar that survives being tested against multiple time-horizons without being retrospectively falsifiable. The regulatory framework wrote the sentence; the company wrote it the way the framework permits.³

On 2026-04-13, Attorney General Ken Paxton's office issued a Civil Investigative Demand to Lululemon USA, Inc. The AG's verbatim statement: "I will not allow any corporation to sell harmful, toxic materials to consumers at a premium price under the guise of wellness and sustainability. If Lululemon has violated Texas law, it will be held accountable."⁴

Lululemon's response, same day: "The health and safety of our guests is paramount, and our products meet or exceed global regulatory, safety, and quality standards. We require all our vendors to regularly conduct testing for restricted substances, including PFAS, by credible third-party agencies to confirm ongoing compliance."⁵

The response to the investigation repeats, near-verbatim, sentence three of the corporate page. The same grammar addressed to the consumer on the website is addressed to the Attorney General in the regulatory filing.

The Five Columns

Stand the paragraph next to the rulebook. The rulebook is five columns on the same day.

The sixth row is the United Kingdom. There is no textile-specific PFAS restriction. The UK PFAS Plan, published 2026-02-03, is a planning document, not a statutory instrument. No textile consultation has opened. A British consumer reading "PFAS-free" on a London high-street rack is reading a claim that has not been tested against a domestic standard, because no domestic standard exists.⁶

California's threshold is 100 ppm total organic fluorine. The EU's threshold is 25 parts per billion PFHxA.

4,000 to 1.

The two numbers do not measure the same thing. California measures elemental fluorine from any source via combustion and ion chromatography. The EU measures one specific short-chain perfluoroalkyl acid via targeted LC-MS/MS. They cannot be added. They cannot be subtracted. A consumer with a calculator cannot reconcile them. They can only be passed separately -- which means a "PFAS-free" claim that has cleared one has not necessarily been tested against the other. The 4,000-to-1 ratio is not a chemistry divergence. It is a jurisdictional divergence masked by identical grammar on the label.⁷

Three laboratory methods exist for measuring PFAS on textiles. Each catches what the others miss.

Total Fluorine sees all fluorine in the sample -- PFAS and non-PFAS alike. It is the broadest net and the bluntest instrument: a proxy screen that, in OEKO-TEX's own stated rationale, exists to restrict intentional PFAS use while not prohibiting all fluorine-containing substances in textiles.⁸ Extractable Organofluorine sees fluorine extractable by solvent at ambient conditions but misses polymer-bound fluorine entirely -- what is inert in extraction today may release in weathering over twelve to twenty-four months. Targeted analysis via EPA Method 1633 or equivalent identifies a pre-defined list of forty specific PFAS molecules; it cannot detect anything not on the list, and the list is always historical.⁹

Lululemon's Restricted Substances List, per published analysis of the RSL 9.0, caps total organic fluorine at 50 ppm on a finished-goods basis -- tighter than California's 2027 threshold on paper, produced by the same proxy methodology OEKO-TEX uses.¹⁰ What the RSL does not publicly disclose: which specific test method, which detection limit per substance, which lab, whether the test is run on finished garments or raw materials, whether the garment is tested after simulated wear. The threshold is published. The methodology behind it is not.

OEKO-TEX's own 2026 provisions allow an exception from the total-fluorine limit if the applicant demonstrates the fluorine source is non-PFAS -- speciation happens downstream of the threshold, by manufacturer-supplied documentation, not by a second test.¹¹

Five jurisdictional columns. Three laboratory methods. One SKU. Fifteen cells in a grid the consumer cannot see. The composition of those fifteen cells into a yes-or-no answer is the consumer's unpaid labour, performed on inputs the consumer is not given.

The Twist

A garment certified "PFAS-free" at factory gate under a targeted-analyte method against a polymer-exempt rulebook can, eighteen months later, exceed the same rulebook's threshold without any chemistry having been added to the garment. The test certified conditions at time zero. The garment ages. The regulation sampled the factory, not the wardrobe.

The premise that most current PFAS restrictions encode: polymer-bound fluorine is inert. Side-chain fluorinated polymers, fluoropolymer binders, PTFE laminate -- these are treated differently from monomeric PFAS in the EU universal restriction's draft derogations and in most US state instruments. The industry-regulatory position is that polymer-bound fluorine is chemically inert and therefore bioinaccessible. The regulatory framework encodes that position as a class exemption.

Schellenberger et al., published in Environmental Science & Technology in 2022, ran a six-month outdoor weathering study on coated textiles. Across all three side-chain fluorinated polymer treatments, the study found "a large (up to 100-fold) increase in PFAA concentrations after weathering."¹² Each finish emitted a distinct pattern of short-chain PFAS. In some cases, post-weathering concentrations exceeded the regulatory limits that had been set on the assumption the polymer would remain bound.

Van der Veen et al., also 2022: 6:2 fluorotelomer alcohol on coated polyamide rose from 87 micrograms per kilogram to 430 -- a 4.9-fold increase -- after aging plus five wash cycles. The OECD's 2022 synthesis on side-chain fluorinated polymers confirms, across multiple studies, that these polymers function as precursors: they hydrolyse, photolyse, and oxidise into short-chain PFAS over the garment's service life.¹³

Someone wrote a rule that sampled time zero and encoded a chemistry premise that aged out. Report 046, The Class Exemption, mapped this structural move in cookware: a class-level regulatory exemption surviving because the regulator treats the class as the unit and regulates individual molecules. Here the same move extends from cookware to performance apparel, with different peer-reviewed chemistry -- Schellenberger's SFP precursors, not the PFOA/PFOS class-level reads of 046. The industry is different. The exemption is the same. The class-wide inertness premise on which it rests has been falsified in the peer-reviewed literature for at least four years, and the exemption has not moved.¹⁴

The method sampled time zero. The garment ages across a service life. No current rulebook samples that.

1,546 Days

On 2022-01-18, Environmental Health News and Mamavation published an investigation. An EPA-certified laboratory tested thirty-two pairs of women's activewear. Eight returned detectable fluorine. Concentrations ranged from 10 to 284 parts per million. Named products included the Lululemon Align Highrise Pant. The study used total fluorine screening at a 10 ppm detection threshold.¹⁵

The same year, Rodgers et al. at the Silent Spring Institute tested ninety-three consumer products marketed to or used by children and adolescents. Fifty-eight percent contained detectable fluorine. In the authors' words: "PFAS were found only in products labeled as water- and/or stain-resistant, regardless of whether they were marketed as 'green' or 'nontoxic.'"¹⁶

On 2026-04-13, the Texas Attorney General issued a Civil Investigative Demand to Lululemon USA, Inc., citing the same chemistry, the same brand, the same product line.

1,546 days elapsed between the published lab finding and the first dated state action.¹⁷

A counter-reading of the lag is available and deserves its full weight. The 2022 studies used screening methodology -- total fluorine -- that cannot by itself distinguish PFAS from other fluorinated compounds. The state attorneys general correctly required confirmatory targeted analysis before issuing investigative demands. The four-year gap is due diligence, not regulatory failure. The consumer-journalism findings informed the regulatory work but could not substitute for it.

The counter stands on the science. What the counter does not address is the architecture: not whether the 2022 screening was sufficient, but whether the system that received the screening had any funded, named, timed pathway to act on it.

An investigative finding in an EPA-certified lab crossed a published threshold, and no funded pathway existed to escalate it to confirmatory testing within the regulatory system. The labs that found the chemistry were funded by donations and subscriptions, not by any state budget line.

The consumer-chemistry investigation of this category was conducted by Mamavation, Environmental Health News, Silent Spring Institute, Toxic-Free Future, and the Environmental Working Group -- organisations funded by subscription, donor contribution, and research grant, not by the brands they test.¹⁸ The consumer's mental model of the verification system is that regulators set standards, accredited labs verify them, brands comply, watchdogs corroborate. The observed sequence is the reverse: donor-funded labs find the chemistry, journalism publishes it, brands restate their present-tense claims, regulators arrive years later. A fourth rail is doing the load-bearing work. That rail runs on a different budget from the one the consumer is shopping in.

The pattern is not confined to textiles.

Medical device conformity assessment -- CE marking in the EU, FDA clearance in the US, UK CA Marking -- produces different verdicts on the same physical device under different evidence standards. Essure, a permanent birth-control implant, had its EU CE mark suspended in August 2017 and was not withdrawn from the US market until December 2018. The same device under concurrent regulatory regimes, reaching different conclusions over a year apart.¹⁹

Food allergen labelling: sesame was a mandatory declared allergen under EU Regulation 1169/2011 from 2014. The United States did not add sesame to its major allergen list until the FASTER Act took effect on 2023-01-01 -- roughly nine years later. The same biological allergenicity under different labelling mandates.²⁰

Sustainable-fund disclosure: in a single quarter, Q4 2022, 307 funds classified as SFDR Article 9 were downgraded to Article 8, with combined assets under management of approximately 175 billion euros -- not because the underlying portfolios changed but because the European Securities and Markets Authority clarified that Article 9 funds must be exclusively composed of sustainable investments. Same assets. Same managers. Same quarter. Verdict changed because the rubric was reread, not because the object changed.²¹

Where three conditions co-exist -- a compressed consumer-facing claim, a plural rulebook with method-level divergence, and a risk time-horizon longer than the regulatory instrument -- the same three-axis gap recurs: space, method, time. The three cases were pre-registered at Stage 7 of the research process; each returned independently; the test was blind to outcome. The claim is about that configuration, not about labels in general. This structural pattern -- five rulebooks, three methods, three time-horizons, one word -- is what this report names The Epistemic Lattice: a certification claim that is singular on the label but plural in the regulatory system behind it, surviving only because the work of composing that plurality is externalised to the consumer who cannot see it.²²

The Levers

Four structural changes, each named from primary regulatory evidence, each with precedent inside the existing framework.

Space-axis closure. A single finished-garment testing standard composing total fluorine, extractable organofluorine, and targeted analysis against a named detection limit, adopted across at least the EU, California, and the UK, with jurisdictional thresholds published in a single public register searchable by consumer. The EU Empowering Consumers Directive 2024 and the ECHA universal PFAS restriction provide the infrastructure. The missing piece is the harmonised finished-garment test.

Method-axis closure. Mandatory disclosure of test methodology, lab identity, detection limit, and targeted-analyte list on any "PFAS-free" or equivalent claim. France's VOC emission labelling regime -- the A+/A/B/C system under Decret 2011-321 -- provides precedent: methodology-transparent, consumer-readable, graded by lab result.²³

Time-axis closure. Post-weathering finished-garment testing at defined wear-cycle intervals, with a mandatory re-certification trigger on the precursor hydrolysis pattern Schellenberger et al. characterised. The van der Veen 4.9-fold increase is not a laboratory curiosity. It is a regulatory parameter that no current rulebook samples.

Leading-indicator formalisation. Statutory recognition of donor-funded third-party testing as admissible leading evidence, with a shortened agency response window when EPA-certified lab findings cross published thresholds. 1,546 days is the number against which any "shortened window" gets measured.

As of the date of this report's research, no CMA enforcement action under the Digital Markets, Competition and Consumers Act 2024 has been opened on textile PFAS claims.²⁴ The CMA has a greenwashing power of up to 10 percent of global turnover. The instrument has not been applied. The UK Parliament's Environmental Audit Committee completed its inquiry into PFAS with a final report published 2026-04-23, following oral evidence sessions from June 2025 through February 2026; the report recommends restrictions on PFAS in non-essential consumer products -- naming food packaging, cookware, and school uniforms -- but activewear-apparel PFAS specifically is not foregrounded.²⁵ The category sits inside the inquiry's silence, not outside it.

Pick up the leggings from the rack on Regent Street. Hold the swing ticket between two fingers. To know what the word on that ticket means, consult five government websites in three languages, request three laboratory analyses at three detection limits, and wait four years and three months for a state attorney general to read the same page you are reading now. No single word on any ticket could ever have held it.

What Would Change This Analysis

Three findings would narrow or overturn the gap this report identifies.

First, convergence. If the ECHA universal PFAS restriction, currently in final RAC opinion (adopted 2 March 2026) with SEAC public consultation open until 25 May 2026, enters into force with a unified textile threshold harmonised with California's AB 1817 schedule and adopted by the UK under its own REACH framework -- the space axis closes. OEKO-TEX's planned reduction to 50 mg/kg total fluorine from June 2026, tracking California's 2027 schedule, is evidence that convergence is in motion.²⁶ The convergence is a future-tense claim. The Lattice is a present-tense observation. Both can be true simultaneously.

Second, methodology transparency. If OEKO-TEX or a successor standards body publishes, per SKU and per certificate, the test methodology used, the lab identity, the detection limit, and the targeted-analyte list in a format searchable by consumers, the method axis closes. The infrastructure to do this already exists within OEKO-TEX's seventeen partner institutes; the disclosure does not.

Third, the RSL. If the Lululemon Restricted Substances List 9.0, when its full contents are publicly accessible, discloses the test methodology, the detection limit per substance class, the finished-garment versus raw-material scope, and a post-weathering re-testing protocol -- the report's characterisation of the claim narrows. As of the date of this research, the publicly accessible portions of the RSL disclose the threshold and the AFIRM alignment but not the methodology behind them.²⁷