Ad index

Supplement ad index: November 2026

Supplement ads running on Meta, the share that carried a line Vera flags, the rules most hit and how long the ads had been running when we read them.

Flagged for review, not a ruling. Meta makes the final call on every ad. Counts only.

Ads read

From 1 to 3 November 2026 we read 192 supplement ads in Meta’s public Ad Library, in four sweeps, one ad from each of 192 brands (counted by the website the ad leads to). Vera, Merchient’s AI employee, read each ad’s words against Meta’s ad rules and the FTC and FDA guidance, and gave it a tier.

• Red. Wording Vera expects a regulator or a platform to act on.
• Yellow. Every other finding, less likely to be acted on.
• Green. Nothing to flag.

146 of the 192 ads (76%) carried a line Vera flags, Red or Yellow.

Not counted: 115 more ads were found by looking up a brand by name, not by a search word. We chose those brands, a search did not find them, so they are left out of every number on this page.

The five rules most hit

Each flagged ad is counted once, under the rule its reason points to; we sorted Vera’s reasons into these groups. These five cover 126 of the 146 flagged ads. The other 20 are spread over five smaller groups. Examples are described, not quoted: no brand, product name or ad is named.

1. FDA: a line that promises relief from, or names, a disease or a symptom. 50 ads, 49 of them Red. The FDA reads a line that says a product relieves, fixes or prevents a disease or a symptom as a drug claim. A supplement can describe how it supports the body; it can’t say it treats a condition. Example: A headline that promises to fix a named gland.
2. FTC: proof or ranking words with no basis shown. 36 ads, 2 of them Red. The FTC expects a brand to be able to prove what its words claim. “Clinically proven”, “science-backed”, “best-selling” and “the only” need evidence behind them, and an ad that shows none is flagged. Example: An ad that calls its product “scientifically proven” and names no study.
3. Meta: a line that tells or asks the viewer about their own health. 21 ads, 16 of them Red. Meta’s personal-attributes rule bars an ad that tells or asks the viewer about their own health, with words such as “you have”, “struggling with” or “diagnosed with”. Example: A line that tells the viewer they have a named urinary symptom.
4. FDA: a word that points to a condition, with no rule named. 10 ads, none of them Red. Vera’s word pass flagged a word that points to a health condition, but the judge named no rule for the line, so a person should read it before it leads an ad. Example: A line with a symptom word in it that only the word pass flagged.
5. FDA: the product set beside a prescription drug, a medical device or surgery. 9 ads, 9 of them Red. Naming a prescription drug, a medical device or surgery next to a supplement, or pitching the supplement to people on a drug, reads to the FDA as a disease claim. Example: A headline that offers a supplement to people on a weight-loss drug.

By category

Ads are grouped by the search word that found them, one group for each ad. “Other” holds the searches that fit none of the groups, such as collagen gummies, anxiety supplement, appetite suppressant and mens multivitamin.

• GLP-1 and weight. 20 ads read, 13 flagged (65%). Most hit, tied at 3 ads each: beside a prescription drug (FDA) and a condition word, no rule named (FDA).
• Gut. 21 ads read, 17 flagged (81%). Most hit: disease or symptom claim (FDA), 8 ads.
• Brain. 7 ads read, 3 flagged (43%). Most hit: speaks to the viewer’s health (Meta), 2 ads.
• Hormones. 30 ads read, 24 flagged (80%). Most hit: disease or symptom claim (FDA), 10 ads.
• Joints. 11 ads read, 11 flagged (100%). Most hit: disease or symptom claim (FDA), 7 ads.
• Other. 103 ads read, 78 flagged (76%). Most hit: disease or symptom claim (FDA), 23 ads.

Small groups: Brain and Joints each have fewer than 15 ads, so read their shares as counts.

How long the ads had been running when we read them

Meta’s library shows the day each ad started. “Days running” is the number of days between that day and the day we read the ad. For each tier the table gives the median and the first and third quartiles of days running, from the days-running file of the ads in this issue.

The days-running file had no row yet for 3 of the 192 ads read; they are counted everywhere else on this page and left out of this table.

What it shows. Of the 81 Red lines we read that have a days-running row, the median ad had been running 96 days; Green: 117.5 days. Correlation only. Running 90 days or longer: Red 52%; Yellow 52%; Green 54%.

What it does not show. These ads were running when we read them, so this cannot say how many flagged ads were stopped, by the brand or by Meta: an ad that had stopped before a sweep is not in these numbers. A recheck of the same ads would show that; this issue does not. It does not say whether Meta has reviewed any of these ads or what it decided. Every sweep took the library’s active ads sorted by impressions, so the ads lean towards those that reach many people. These ads are not a random sample of supplement ads.

How long lines run, by category and tier

For each category and tier: the median number of days the ads had been running when we read them, and the first and third quartiles (a quarter of the ads had run less than the first quartile, a quarter more than the third). Quartiles are found by linear interpolation between ranks. A figure is shown only for a category and tier with at least 8 ads.

Correlation only: how long ads ran, not whether they sold or were reviewed.

Too few ads to show a figure (fewer than 8; the number of ads is after each tier). Sleep: Red 5, Yellow 0, Green 3; Gut: Green 6; Immunity: Red 2, Yellow 7, Green 2; Brain: Red 4, Yellow 7; Joints: Yellow 2, Green 0; Skin and hair: Red 6, Yellow 4, Green 2.

This table counts 180 ads by category and tier, computed 3 November 2026. 137 of them are the ads on the Winning lines pages, read 30 October 2026 to 2 November 2026. Its counts are not the counts in the tables above, which are for the 192 ads of this issue.

What changed since last issue

This is the first issue, so there is nothing to compare with yet.

How Vera decides

Vera gives each ad one tier: Red for wording she expects a regulator or a platform to act on, Yellow for any other finding, Green for nothing to flag. Each finding carries the rule it rests on and is marked checked or not checked. See a finding laid out in the sample report.

Method

We read supplement ads in Meta’s public Ad Library: active ads for the United States, sorted by impressions, found by search words such as the ones grouped above, one ad from each brand. Vera, an AI, read each ad’s words against Meta’s published ad rules and the FTC and FDA guidance. A tier and a rule show what Vera flagged. We do not know what Meta decided about any of these ads. Vera read each ad’s headline and the start of its text, not its picture or video or the page it leads to, so a Green ad had nothing to flag in those words and may still have a problem elsewhere. Start dates are the ones the library shows.

Which lines ran longest, each with Vera’s tier and rule: What runs longest: Brain supplement ads; What runs longest: GLP-1 and weight supplement ads; What runs longest: Gut supplement ads; What runs longest: Hormones supplement ads.

Want to know where your own ads fall? Put your website in the box on the home page and see your own panel.

Risk review, not legal advice.

All issues · Weekly · Back to the home page