Ad Claims Compliance: Substantiation for Marketers
Every claim in an ad needs evidence behind it before it runs. Here is a practical guide to substantiation, risky claim types, and a review process that keeps ads approved.
Most ad rejections and most regulatory trouble come from the same root cause: a claim that the advertiser could not back up. "Clinically proven." "Lasts twice as long." "Number one rated." "Results in seven days." Each of these is a factual statement, and each needs evidence that existed before the ad ran.
That idea, that claims must be supported in advance, is called substantiation. It is the core of ad claims compliance in the US and in many other markets, and ad platforms like Meta and OpenAI build their own rules on top of it.
This guide is practical rather than legal. It explains what substantiation means for a working marketer, which claims carry the most risk, and how to set up a simple review process. For anything high-stakes, especially health, finance or children's products, talk to a lawyer who knows your category and market.
What substantiation means
In the US, the FTC's long-standing position is that advertisers must have a reasonable basis for objective claims before they make them. The level of evidence required depends on the claim. A claim like "clinically proven" implies clinical evidence. A claim like "made in small batches" implies records showing it.
Three ideas help in practice:
- Claims are judged by their net impression. What would a reasonable consumer take away? If your ad strongly implies a health benefit, it is treated as a health claim even if you avoid the exact words.
- Evidence must come first. You cannot run a claim and look for support later.
- The evidence must match the claim. A study on one ingredient does not automatically support a claim about your finished product.
Ad platforms layer on their own rules. OpenAI's ad policies, for example, prohibit deceptive or unverifiable claims, invented statistics, fake reviews and guaranteed outcomes. Meta's advertising standards similarly prohibit misleading claims and restrict several sensitive categories.
Types of claims, from low to high risk
| Claim type | Example | Risk | What you need |
|---|---|---|---|
| Puffery | "Our coziest sweater yet" | Low | Generally fine if obviously subjective |
| Product facts | "100% merino wool" | Low if true | Spec sheets, supplier documentation |
| Process claims | "Roasted within 48 hours of shipping" | Low to medium | Records showing it is routine |
| Performance claims | "Keeps drinks cold for 24 hours" | Medium | Your own testing, documented |
| Comparative claims | "Lasts longer than leading brands" | High | Head-to-head testing, fair comparison |
| Superlatives | "The best-selling serum" | High | Market data covering the claim exactly |
| Health and wellness | "Reduces joint pain" | Very high | Competent and reliable scientific evidence; often prohibited outright |
| Income or results | "Make $5,000 a month" | Very high | Typically prohibited on ad platforms |
Puffery is the one area where vague praise is allowed, because no one takes it as a measurable fact. But the line is thinner than people think. "The best" is often puffery. "Rated best by customers" is a factual claim that needs data.
The claims that cause the most trouble
"Clinically proven" and "scientifically tested"
These imply clinical or scientific evidence on your product. If the study was on an ingredient, at a different dose, or in a lab rather than in people, the claim likely overstates it.
Numbers and percentages
"93% saw smoother skin" needs a study that measured exactly that, with a method that would hold up. Invented or loosely extrapolated statistics are explicitly banned in ChatGPT ads.
Superlatives
"Number one," "best-selling," "most effective" all need data that covers the market and time period implied. Without it, rephrase into something specific and true.
Comparisons to competitors
Comparative claims need fair, current, like-for-like evidence. In ChatGPT ads you cannot use other companies' trademarks at all, which rules out named comparisons.
Health, weight and wellness outcomes
Health, supplements, wellness and weight management are restricted categories on ChatGPT ads, and guaranteed health outcomes and before-and-after transformations are prohibited. Meta has its own restrictions here. If you sell in these categories, read our guide to ChatGPT ads for supplements and wellness.
Testimonials
A testimonial is treated as a claim by the advertiser. If a customer says "this cured my eczema," you are now making that claim. Testimonials also need to reflect typical results. See how to use social proof without breaking rules.
Urgency and scarcity
"Only 3 left" and "Sale ends tonight" are factual claims. If they are not true, they are deceptive. Fake urgency is prohibited in ChatGPT ads and is a frequent consumer protection target.
Rewriting risky claims
You can often keep the persuasive point while removing the unsupported claim. Specific, true details are usually more convincing anyway.
| Risky | Safer, specific |
|---|---|
| "Clinically proven to reduce wrinkles" | "Contains 0.5% retinol in a fragrance-free base" |
| "The best dog bed ever made" | "Ballistic nylon cover with a 90-day chew warranty" |
| "Lose 10 pounds in a month" | Usually not fixable; reframe around a non-outcome benefit, like recipes or convenience |
| "Twice as durable as other brands" | "Tested to 50,000 abrasion cycles in our lab" (if you have the test) |
| "Thousands of 5-star reviews" | "Rated 4.6 by 2,100 verified buyers" (exact, current) |
| "Only a few left!" | "Ships by Friday if ordered today" (if true) |
A substantiation file for every ad
The simplest compliance habit is to keep a file for each claim before the ad launches. For each claim, record:
- The exact wording in the ad
- What a reasonable reader would take it to mean
- The evidence that supports it (document, test, data export)
- The date the evidence was gathered and when it should be rechecked
- Who approved it
A shared spreadsheet with links to evidence documents works fine. When a platform reviewer or a customer challenges a claim, you can answer in minutes.
A pre-launch review checklist
Run every ad through these questions before it goes live:
- Does every factual claim have evidence on file?
- Does the overall impression imply anything you cannot support?
- Are numbers exact, current and sourced?
- Are testimonials genuine, permissioned and typical?
- Are paid relationships disclosed?
- Does the ad avoid competitor names and trademarks?
- Does the ad avoid health cures, income guarantees and before-and-after images?
- Is any urgency real?
- Does the landing page support every claim in the ad?
- Is the product or category restricted on this platform?
For ChatGPT ads, the ChatGPT ad policy checker flags common problem phrases, and our full ChatGPT ad policies guide covers what is allowed, restricted and banned.
Platform-specific notes
Meta
Meta reviews ads automatically and sometimes manually, and checks the landing page as well as the ad. Common rejection reasons include personal-attribute language ("Are you struggling with debt?"), exaggerated health claims and misleading before-and-after images. Repeated rejections can lead to account restrictions, so fix root causes rather than resubmitting slightly reworded ads. Our guide on why Meta ads get rejected goes through the common cases.
ChatGPT ads
OpenAI's ad policies, as of 2026, prohibit deceptive and unverifiable claims, fake reviews, invented stats, guaranteed outcomes, before-and-after transformations, exploiting hardship or insecurity, implying the ad knows what the user said, impersonating OpenAI or other brands, and using other companies' trademarks. Health, financial products, alcohol, dating, weight management and products for minors are restricted. Ads must match their landing pages.
Because ChatGPT ads appear right below an answer, they also carry a trust burden. A modest, specific claim reads as helpful. An inflated one reads as out of place.
A hypothetical compliant example for a supplement-adjacent product, written with the restrictions in mind:
- Headline (39 chars): Magnesium glycinate, third-party tested
- Body (88 chars): 120 capsules per bottle, no fillers. Batch test results published on every product page.
Note it makes no health outcome claim at all. The proof is verifiable product facts.
Building compliance into the workflow
Compliance works best as a step in the process, not a final gate.
- Brief. List approved claims and their evidence at the start of each creative brief.
- Write. Writers use only approved claims, or flag new ones for review.
- Review. A second person checks the ad against the checklist.
- Launch. The substantiation file is linked from the ad record.
- Recheck. Review claims quarterly, or when the product, formula or data changes.
How SecondWin handles claims
Every ChatGPT ad SecondWin writes is checked against OpenAI's ad policies and SecondWin's own rules, which block unproven claims, competitor names and made-up facts. It writes only from proof it can find on your site, so if a claim is not supported on your landing page, it does not go into the ad. That keeps message match tight and lowers the chance of rejection.
You can see the claims and messages SecondWin would use with a free URL analysis. Plan details are on the pricing page.
FAQ
What is ad substantiation?
Ad substantiation means having evidence that supports a factual claim before the ad runs. The kind of evidence depends on the claim: product specs for a materials claim, documented tests for a performance claim, and competent scientific evidence for a health claim. In the US, the FTC expects advertisers to have a reasonable basis for objective claims, and ad platforms enforce their own versions of the same principle.
Is puffery allowed in ads?
Generally yes, when it is clearly subjective and no reasonable person would take it as a measurable fact, such as "our coziest sweater." Puffery becomes a problem when it implies something measurable, like "rated best by customers" or "the most effective formula." When in doubt, replace the superlative with a specific, true detail, which is usually more persuasive anyway.
Can I make health claims in ChatGPT ads?
Health, supplements, wellness and weight management are restricted categories on ChatGPT ads, and guaranteed health outcomes and before-and-after transformations are prohibited. Any claim needs strong evidence and must match the landing page. Many brands in these categories do better focusing ads on verifiable product facts, such as ingredients, testing and sourcing, rather than outcomes.
What happens if my ad makes an unsupported claim?
At minimum, the ad may be rejected. Repeated violations can lead to account restrictions on Meta or ChatGPT. Beyond platforms, unsupported claims can draw complaints, chargebacks and, in some cases, regulatory action. The cheapest protection is a simple substantiation file and a pre-launch checklist so every claim is backed before it runs.