When a Cosmetic Becomes a Drug: The Marketing Language Trap Under 21 CFR Part 201

The ingredient list didn’t change. The formulation didn’t change. But one sentence on the product webpage — “clinically proven to stimulate collagen synthesis and restore skin’s natural repair function” — turned a $45 facial serum into an unapproved new drug. FDA sent the warning letter six months later.

That scenario plays out dozens of times a year across skincare, personal care, and supplement brands. Marketing teams, optimizing for conversion, choose language that resonates with consumers but quietly crosses a line the Federal Food, Drug, and Cosmetic Act has held firm since 1938. The line isn’t about what’s in the bottle. It’s about what you said the bottle does.

The Legal Line FDA Draws — and Most Marketing Teams Don’t Know Exists

The FD&C Act defines a drug, at Section 321(g)(1), as any article “intended to affect the structure or any function of the body of man.” A cosmetic, under Section 321(i), is intended to be “rubbed, poured, sprinkled, or sprayed on, introduced into, or otherwise applied to the human body for cleansing, beautifying, promoting attractiveness, or altering the appearance.”

Notice what those definitions don’t overlap on: function. A cosmetic alters appearance. A drug affects structure or function. The moment your marketing copy claims the latter — even parenthetically, even in fine print, even in a customer testimonial you’ve approved — FDA’s “intended use” doctrine can apply.

Intended use isn’t self-declared. FDA determines it from labeling, advertising, promotional materials, and verbal statements made by company representatives. Under 21 CFR 201.128, “labeling” includes anything accompanying the product — which courts have interpreted broadly to include websites, social media, and third-party retailer listings you’ve provided content for. If your Amazon A+ page says the product “reduces inflammation at the cellular level,” you’ve labeled it as a drug, whether or not your regulatory team ever saw that page.

The practical consequence is significant. A cosmetic that makes drug claims must comply with the full pharmaceutical compliance stack: 21 CFR Parts 210 and 211 for GMP, a New Drug Application or Abbreviated NDA through CDER (unless it qualifies under the OTC monograph system), and labeling satisfying 21 CFR Part 201 — including adequate directions for use and complete Drug Facts. Very few cosmetic manufacturers are set up to do any of that.

The Words That Flip the Switch

Knowing that intended use is the trigger, the practical question is: which words actually flip the classification? FDA’s enforcement history and published guidance give us a fairly clear picture.

Structure/function language is the highest-risk category. Phrases like “rebuilds collagen,” “regenerates skin cells,” “restores the skin barrier at the molecular level,” or “activates the skin’s natural repair mechanisms” all describe changes to how the body works — not just how it looks. These are drug claims under the doctrine, full stop.

Disease references elevate risk further. Any mention of treating eczema, rosacea, psoriasis, acne (beyond “controls acne” under the approved OTC monograph), dermatitis, or wound healing moves a product from cosmetic into drug territory almost automatically. Even “soothes redness associated with eczema” can be enough. The FDA has consistently held that referencing a disease condition implies a therapeutic intent.

Comparative physiology language is subtler but equally dangerous. Statements like “your skin produces less hyaluronic acid as you age — our formula restores it” describe a corrective action on a physiological process. That’s a structure/function claim dressed up as consumer education.

Contrast those with clearly cosmetic claims: “hydrates and softens skin,” “minimizes the appearance of fine lines,” “leaves skin looking more radiant,” “improves skin tone.” The difference is between affecting how skin looks and affecting how skin works. One word — “appearance” versus “function” — carries enormous legal weight.

And some categories carry settled classifications regardless of marketing. Sunscreen products with SPF 4 or higher have been OTC drugs since FDA finalized its sunscreen monograph in 1978. Fluoride toothpaste is a drug. Anti-dandruff shampoo containing zinc pyrithione or selenium sulfide is a drug. Antiperspirant is a drug (while plain deodorant isn’t). These dual-category products must comply with their applicable OTC monograph requirements, and any claims beyond what the monograph permits require FDA approval.

The Compliance Consequence Is Not Small

Let’s be direct about what happens when FDA concludes your cosmetic has been marketed as a drug without approval. It doesn’t suggest you clean up the website.

An unapproved new drug on the market is a per se violation of 21 U.S.C. § 355. FDA can issue a Warning Letter demanding formulation or label changes and market withdrawal, refer the matter to the Department of Justice for injunction or seizure, and place the company on Import Alert — which blocks all shipments at the border until remediation is verified. For manufacturers currently operating under cosmetic GMP (FDA’s 21 CFR Part 700 voluntary guidance), the jump to pharmaceutical GMP under 21 CFR Parts 210 and 211 can cost anywhere from $250,000 to several million dollars in facility upgrades, documentation overhaul, process validation, and personnel training.

That’s before accounting for reputational damage. Warning letters are public within days of issuance and are fully searchable on FDA’s website. Retailers, Amazon, and distributors actively monitor that database, and many carry contractual clauses allowing them to pull products or terminate agreements on FDA action. Recovery from an import alert can take 12 to 24 months even after the underlying issue is resolved.

The typical warning letter response cycle — receiving the letter, retaining regulatory counsel, preparing a response, completing remediation — runs 3 to 9 months and routinely costs $50,000 to $200,000 in legal and consulting fees alone, not counting manufacturing or relabeling. None of that is recoverable. And nearly all of it was preventable at the marketing review stage.

How AI-Augmented Content Review Changes the Risk Equation

Here’s the operational problem: the people who write marketing copy — copywriters, brand managers, performance marketers — are optimizing for conversion, not regulatory classification. And regulatory affairs teams rarely see every piece of content before it goes live. The review process, where it exists, is typically episodic: someone flags something, a reviewer looks at it, maybe it gets revised. There is no systematic screening of every webpage, social post, distributor brief, or trade show card.

That’s exactly the gap AI-augmented regulatory review is built to close.

The approach is straightforward in concept: train a model on FDA’s published body of warning letters (CDER and CFSAN have made hundreds publicly available) and on the statutory language of the FD&C Act and 21 CFR Part 201, then run marketing content through that model to score each phrase against known trigger patterns. The system doesn’t make legal determinations — it shouldn’t, and a well-designed one won’t pretend to. But it surfaces language for human regulatory review that would otherwise disappear into a 500-page content library.

A trained content audit system can process an entire product catalog’s marketing assets in hours, flagging specific phrases with confidence scores and citations to analogous FDA enforcement examples. We’ve seen this deployed most effectively in pre-launch content audits, where a brand runs its full planned marketing suite through AI review before a single product page goes live. Issues that would have required a Warning Letter to surface get caught in a structured spreadsheet instead. That’s the intervention point that actually protects the business.

Several outputs matter in practice: a structured log of flagged phrases with regulatory rationale, a priority ranking (high, medium, or low risk based on enforcement frequency), and suggested alternative language that preserves the marketing intent without crossing the statutory line. That last step — suggesting “maintains the appearance of youthful skin” as a safe substitute for “restores skin’s youthful function” — is where AI tooling and regulatory expertise work best together. The model surfaces the pattern; the human owns the recommendation.

What Manufacturers Should Do Before the Next Product Launch

If you haven’t run a systematic marketing content audit under the intended use doctrine, start with your highest-traffic digital assets: main product pages, Amazon listings, your top-performing email campaigns, and any influencer content you’ve approved or provided copy for. These are the assets FDA will review if your brand draws attention.

For each piece of content, apply one filter: does this claim describe how the product changes the appearance of the body, or how the body functions? If it’s the latter, it needs regulatory review before it stays live — not after.

Second, build pre-publication review into the brief stage, not the final draft stage. Regulatory input on claim vocabulary before copywriters start writing is orders of magnitude cheaper than retroactive remediation. A 30-minute claim strategy session upstream saves 30 weeks of crisis management downstream.

Third, map your product portfolio against the OTC monograph system. If any products contain active ingredients covered by a finalized monograph — sunscreen, anti-dandruff, acne, skin protectant, oral care — confirm that labeling is fully monograph-compliant and that no marketing asset makes claims outside the monograph’s permitted scope.

The language your marketing team uses today is the evidence FDA will exhibit in an enforcement action next year. Treating label and content review as a one-time event rather than a continuous process is how companies pay $150,000 to fix something that would have cost $5,000 to catch at the source.

Written by Sam Sammane, Founder & CEO, Aurora TIC | Founder, Qalitex Group. Learn more about our team

Reserve early access to our AI audit tools — including AI-powered marketing content review for drug/cosmetic classification risk. Contact us

Related from our network

• ISO 17025 Laboratory Testing for Label Claims and Product Compliance — Qalitex Laboratories provides accredited testing support to help manufacturers substantiate cosmetic and supplement claims with data FDA can scrutinize.
• Health Canada NHP and Cosmetic Classification for Canadian Manufacturers — Androxa supports Canadian brands navigating Health Canada’s parallel distinction between cosmetics, natural health products, and drugs under federal regulations.