What a supplement or functional beverage brand is not allowed to say.

Three rulebooks, and almost everyone treats them as one.

A founder who has just had an ad rejected usually asks a single question: is this legal? That is the wrong question, because three separate authorities are looking at three different things and only two of them are the law.

The order matters, because the consequences arrive in reverse. Almost nobody meets the FTC. Almost everybody meets an ad rejection. And the rule that produces most of those rejections is not a law at all, it is one line in Meta’s policy that has nothing to do with health claims. It is in Part 2 below.

One thing to settle before anything else. Complying with FDA does not protect you from FTC. The FTC says so in its own guidance: “The FDA has primary responsibility for claims that appear in labeling… The FTC has primary responsibility for claims in all forms of advertising.” A label that has been through regulatory review can still carry an ad campaign that the FTC treats as deceptive.

The FTC decides whether you can prove it, not whether it is true.

The governing document is the Health Products Compliance Guidance, issued in December 2022. It replaced the 1998 dietary supplement guide and now covers every health product, supplements and functional foods included.

The standard it sets is competent and reliable scientific evidence, which the guidance defines as tests and studies “conducted and evaluated in an objective manner by experts” and “generally accepted in the profession to yield accurate and reliable results.” In practice that means human trials. The guidance is direct about it: “Randomized, controlled human clinical trials (RCTs) are the most reliable form of evidence and are generally the type of substantiation that experts would require for health benefit claims.” Animal and laboratory work “without confirmation by human RCTs… aren’t sufficient.”

Four consequences of that standard catch brands out, and all four are ordinary practice in the category.

• An ingredient study does not carry your formula. The dosage and the formulation have to match the product that was studied. A blend containing a studied ingredient at a fraction of the studied dose is not supported by that study.
• You cannot select the study that agrees with you. The guidance requires evidence “largely consistent with the surrounding body of evidence.”
• A disclaimer does not rescue a testimonial. Testimonials reporting results more dramatic than users can generally expect “are likely to be deceptive,” and attempts to fix that with “Results not typical” “don’t cure the deception.”
• The DSHEA disclaimer is not a shield. The guidance says plainly that it “won’t cure an otherwise deceptive ad.” The small print at the bottom of the page does not license the headline at the top.

Liability is not confined to the brand either. The guidance reaches ad agencies, retailers, distributors and expert endorsers.

Why the penalty number changed, and who is already exposed to it.

On 13 April 2023 the FTC sent a Notice of Penalty Offenses to, in its own words, “approximately 670 companies involved in the marketing of OTC drugs, homeopathic products, dietary supplements, or functional foods.” The notice is not an accusation. It is a delivery mechanism: it establishes that the recipient knows which practices have already been found unlawful, and that knowledge is what unlocks civil penalties.

The three practices the notice names are worth reading as written, because they are the substantiation rules restated as offences:

The 2023 announcement put the exposure at up to $50,120 per violation. That number is now out of date and the current one is higher. The maximum published in the Federal Register on 17 January 2025 rose to $53,088, and the FTC was still using that figure in December 2025. Per violation, not per campaign.

Two more FTC instruments apply to anyone selling this category online. The Endorsement Guides, finalised 29 June 2023, close the obvious workaround: an endorsement “can’t be used to make a claim the marketer of the product couldn’t legally make.” Hiring a creator to say the thing you are not allowed to say does not change the rule. And the Reviews and Testimonials Rule, effective 21 October 2024, carries the same per violation penalty and covers fake reviews, undisclosed insider reviews, incentives conditioned on a positive rating, review suppression, and bought followers. The FTC warned ten companies under it in December 2025, so it is being used.

Enforcement is live in this category right now. In April 2026 the FTC took action over “The Only Supplement Clinically Proven to Help Height Growth,” alongside employee written five star reviews. In June 2026 it sued a supplement company over claims its products treat depression, anxiety and ADHD in children. Both are on ftc.gov and both are recent.

The FDA rule that turns a marketing sentence into a drug claim.

A supplement may make a structure and function claim. It may not make a disease claim. The line between them is drawn in 21 CFR 101.93, and paragraph (g) is the part that surprises people, because it does not only look at your sentence.

Paragraph (g) treats a claim as an implied disease claim based on the product name, the formulation, the citations you include, the disease terminology you use, the imagery on the page, and whether the product is positioned as a substitute for a therapy. You can trip it without writing a single claim. A product named after a condition, or a page carrying a study about that condition, has already made the claim.

Making a structure and function claim also carries two obligations that are often missed: the disclaimer, verbatim, that the statement “has not been evaluated by the Food and Drug Administration” and that the product “is not intended to diagnose, treat, cure, or prevent any disease”, and a notification to FDA within 30 days of first marketing the claim.

The warning letters show how ordinary the offending sentences are. A January 2026 letter turned on “Helps regulate blood sugar” and “May relieve the inflammatory pain associated with osteoarthritis.” A December 2025 letter turned on “promoting healthy bone development and height increase.” None of those reads like a drug claim to the person who wrote it.

The functional beverage trap, and it is the most expensive one here.

This is the rule that catches energy drinks, functional sodas and anything else in a can, and it is not a marketing rule. It decides which body of law your product sits under.

The FDA’s guidance Distinguishing Liquid Dietary Supplements from Beverages, final in January 2014, says it in one sentence: “Even when the label of a product characterizes it as a dietary supplement, the product may not in fact be a dietary supplement.” What decides it is how the product presents and sells. The factors named include the product name, whether it uses words like drink, water or soda, the packaging, and the guidance is specific enough to name a “twelve-ounce pop-top aluminum can.” Serving size against normal daily fluid intake counts. So do directions that suggest quenching thirst. So does your marketing, including your metatags and where the product sits on a shelf.

The consequence is the expensive part, and the guidance states it directly: “a dietary ingredient that is lawful for addition to dietary supplements may or may not also be lawful for addition to conventional foods.” A beverage needs food additive approval or GRAS status for what is in it. A supplement uses the new dietary ingredient pathway instead. Calling the can a supplement does not move it into the supplement rules. The way you sell it can move it out of them.

Caffeine sits in the same trap. The GRAS listing at 21 CFR 182.1180 covers caffeine at 0.02 percent in cola type beverages, not in beverages generally, and the FDA says on its own consumer page that it “does not have a regulation specific to ‘energy drinks’.” Added caffeine has to appear in the ingredient list.

And a phrase to stop using in pitch decks. The FDA does not approve new dietary ingredients. Notification under 21 U.S.C. 350b is required at least 75 days before launch for anything not marketed in the United States before 15 October 1994, but a filed notification is not a clearance and advertising it as one is its own problem.

What the four platforms actually say, side by side.

Platform rules are not law and are not obliged to match it. They are stricter than the law in places, looser in others, and they change without notice. Everything in this table was read on the live policy pages on 5 September 2026.

Two rows deserve attention because they contradict what the category believes. Meta’s live policy does not contain the before and after prohibition that almost every agency article attributes to it, and it affirmatively permits before and after transformation imagery for cosmetic products at 18+. TikTok, meanwhile, bans before and after everywhere, in every market, for everyone. The received wisdom has these two platforms the wrong way round.

The rule that rejects more supplement ads than the FTC and FDA combined.

It is not a health rule. It is Meta’s personal attributes policy, and it exists to stop advertisers implying that they know something private about the person seeing the ad. It applies to health, and it does not care whether your claim is true or substantiated.

The policy says ads must not “assert or imply” a person’s attributes, and the list includes “physical or mental health (including medical conditions).” Meta publishes its own examples, and the pattern is unmistakable once you see it.

The difference is the second person. Describing what you sell is allowed. Addressing the reader as though you know their condition is not. That is why the standard direct response opening, the question that names the reader’s problem back to them, fails in this category while the same offer stated as a fact passes. “Struggling with brain fog?” is a violation. “A new approach to focus” is not.

Most brands rewrite the claim when an ad is rejected. The claim was usually not the problem.

The rest of Meta’s health policy, briefly. Dietary, health, weight loss and weight gain products must be targeted to people 18 or over. But the age gate does not apply to “general food products, including protein products” or products “marketed for general wellbeing.” For a functional beverage this is a real fork: sold as a food, no age gate. Sold as a dietary supplement or for weight loss, age gate. Separately, Meta prohibits “statements of inferiority about physical appearance”, close ups pinching fat, claims to cure a named list of incurable conditions, and “promises of specific outcomes within a set timeframe without disclaimers or qualifiers.”

Google names the products. TikTok bans the format. Amazon runs two rulebooks.

Google Ads is the only platform that publishes a list. Its healthcare and medicines policy bars herbal and dietary supplements with active pharmaceutical or dangerous ingredients, anything implying prescription drug equivalence, products subject to a government warning, ephedra, and hCG for weight loss. Behind an accordion on a second page sits a list of roughly 130 named products, mostly slimming teas, sexual enhancement pills and designer prohormones. Two things worth knowing: no Google certification exists for ordinary supplements, so anyone selling you one is selling you nothing, and Google gives at least seven days warning before suspending an account under this policy.

Google is also the platform that states the testimonial rule most usefully: “If you guarantee certain results, you’re required to have a clear and easily accessible refund (money-back) policy. Testimonials that claim specific results must include a visible disclaimer stating that there is no guarantee of specific results and that the results can vary.”

TikTok has no supplement policy at all. Supplements are governed by its weight management, healthcare and misleading content policies instead. Its weight management policy requires 18+ targeting for any weight loss or muscle gain claim and bans “claims that the product alone, without diet or exercise, can lead to weight loss or muscle gain.” Its body image rules are the strictest of the four, prohibiting any suggestion “that there is an ideal body type” or that appearance determines “confidence, self-esteem, emotional wellbeing, desirability, popularity, or social standing.” And its ban on before and after comparisons is global and unconditional. One more that matters now: AI generated content is allowed only if labelled, and undisclosed AI content is rejected.

Amazon is two different problems. To advertise, its healthcare policy sorts supplements into sensitivity tiers. General wellness, sleep and digestion sit in the permissive tier. Weight loss sits a tier down. Fat burners, appetite suppressants, GLP-1 supplements and testosterone or libido products sit in the most restricted tier, where whole surfaces are closed and, critically for anyone running retargeting, building an audience from views or purchases of those products is prohibited outright. Amazon also bans before and after imagery for weight management, guarantees of permanent loss or a specific rate of loss, and requires comparative and endorsement claims to be substantiated with a source dated within 24 months.

Selling on Amazon is the separate problem, and it is the harder one. Amazon’s seller side dietary supplement policy now requires third party cGMP certification for all supplements, with private, first party and consulting audits explicitly refused, and additional heavy metal, microbial and adulterant testing for bodybuilding, joint health, sexual enhancement, sports nutrition and weight management products. Documents are no longer submitted to Amazon directly; an authorised testing and certification provider submits on your behalf. Because this policy is inside Seller Central and two of our own researchers reached different pages for it, confirm the current requirement inside your own Seller Central account before acting on any version of it, including this one.

Twelve things founders believe are fine.

Every row here is something we have seen written in good faith by someone who thought they were being careful.

Three things changed in the last year, and one of them changed yesterday.

California. AB 2030 was presented to the Governor on 4 September 2026. It would prohibit selling weight loss or muscle building supplements to anyone under 18, with a $500 penalty per violation. New York already has this, under a law signed in October 2023. If you sell either category into California, this is the item to watch this month, and the fix is an age gate at checkout rather than anything in your advertising.

Self affirmed GRAS may be ending. The FDA’s 2026 priority deliverables include a proposed rule requiring GRAS notices to be submitted to the FDA for all new substances claimed as GRAS. Today a company can conclude on its own that an ingredient is generally recognised as safe and use it. If that changes, it lands on functional beverage formulation directly, because the beverage pathway is exactly where self affirmed GRAS is used. The same document lists final new dietary ingredient guidance for 2026 and industry best practice for labelling added caffeine.

The Reviews Rule is being enforced. It took effect in October 2024 and in December 2025 the FTC warned ten companies under it. A rule with a $53,088 per violation ceiling and a live enforcement record is not a formality.

The FDA also held a public meeting on 27 March 2026 reopening the question of what counts as a dietary ingredient under DSHEA at all. That one is early, and worth watching rather than acting on.

Five things we could not verify, including two that everybody repeats.

This section exists because a reference that only tells you what it found is not one you can rely on. Each of these was searched for on the live pages and was not there.

• Meta’s ban on before and after images. Very widely cited. We searched the full rendered text of Meta’s health and wellness policy and the Business Help Center version of the same policy and found no such sentence. The page instead permits “general cosmetic products, procedures, surgeries depicting before and after transformation” at 18+. Treat the ban as legacy language until Meta republishes it.
• Meta’s ban on idealized body imagery. Same search, same result. The nearest live equivalents are the ban on statements of inferiority about appearance and the principle against generating negative self perception. TikTok has the explicit rule. Meta does not.
• Meta’s Unrealistic Outcomes policy. The URL still appears in search results and still gets cited. It returns a 404. The live successor is the unacceptable business practices policy.
• A Certificate of Analysis from an ISO 17025 lab as a general Amazon requirement. The phrase does not appear on the live seller policy page. ISO 17025 appears there only in the substantiation path for a non-GMO claim.
• Amazon’s effective date for extending cGMP certification to all supplement categories. Trade press names January 2026. We could not confirm it from any Amazon source. What is confirmable is that the requirement is on the live page now and that Amazon ran an implementation session in December 2025.

One more, from our own benchmark work rather than from policy: the Meta cost per purchase figure for vitamins and supplements that circulates at $45.62 traces back to a vendor whose benchmark pages no longer resolve. We do not use it and neither should you.

How this page was made, and what would make it wrong.

Every rule above was read on the publisher’s own page on 5 September 2026. Where a page is rendered by JavaScript, which is true of Meta’s transparency centre and of Amazon’s help pages, the text was taken from a real browser rather than from a fetch, because a fetch of those pages returns the shell and no policy text at all. Two independent researchers covered overlapping ground and disagreed in one place, on whether Amazon’s seller policy was reachable. Where they disagreed, the page says so rather than picking the more convenient answer.

What would make this wrong. Platform policies change without announcement and without a version history, so any of the platform rows can be out of date the week after publication. Regulator guidance moves more slowly but the penalty figure is adjusted for inflation every January. And a policy read is not a legal read: this page tells you what the published rules say, not whether your specific product and your specific sentence comply with them.

This is a reference, not legal advice. Nothing here is a substitute for regulatory counsel, and for anything with real exposure, a disease adjacent claim, a new ingredient, a beverage that might be a food, you want a lawyer who does this rather than an agency that reads the rules carefully.