Three in four companies on an FTC notice list still publish claims that would face a substantiation question

What we did, in one sentence: we resolved the Federal Trade Commission's five published Notice of Penalty Offenses recipient lists to live websites, read 300 of those websites, and counted what they currently claim.

Nobody has published this measurement before, because until now the two halves did not exist in the same place. The Commission's rosters are public — as five separate PDFs of scanned company names. What those companies assert on their own pages today is public too. Putting the two together is the whole of the method, and the whole of the finding.

Being on one of these lists is not an indication that a company has done anything wrong. That is the Commission's own sentence and it governs everything below. A Notice of Penalty Offenses is not a complaint, a finding, a settlement or an allegation against the recipient. The FTC sent these notices broadly, across whole industries, so that civil penalties of up to $50,120 per violation would become available later against any company that went on to break the law. Nothing here says any company broke it.

What we found

Of 300 recipient websites read:

| | | |---|---| | Readable | 240 (80%) | | Publish at least one assessable claim | 213 of 240 (89%) | | Publish at least one claim that would face a substantiation question | 180 of 240 (75%) | | Claims found in total | 1,867 |

The distribution matters more than the average, because the average is dragged by a long tail:

| Claims per readable site | | |---|---| | 25th percentile | 2 | | Median | 5 | | 75th percentile | 11 | | 90th percentile | 19 | | Maximum | 61 |

Claims that would face a substantiation question ran at a median of 2 per site, and 10 at the 90th percentile.

By which notice the company received

| Notice | Sites read | Publish claims | Median claims | |---|---|---|---| | Substantiation of product claims | 68 | 90% | 8 | | Endorsements and testimonials | 72 | 85% | 6 | | Money-making opportunities | 98 | 90% | 8 | | For-profit education | 11 | 100% | 12 |

The category the Commission wrote to a company about barely predicts how much it publishes. Companies noticed about endorsements publish product claims at almost the same rate as companies noticed about substantiation. Whatever the notice was about, the pages look similar.

What "would face a substantiation question" means, and what it does not

It means a published sentence asserts something a reader would reasonably expect evidence for — an outcome, a comparison, a measurable performance figure, a health effect, an earnings level — and no evidence is cited beside it on the page.

It does not mean the claim is false. It does not mean the company lacks evidence; a substantiation file can sit in a drawer and satisfy a regulator perfectly while never appearing on the website. It is not a legal conclusion and it is not advice. It is a reading of what a page says and what sits next to it, which is all it has ever claimed to be. Our method is published in full, including its limits.

Limits, stated plainly

• The sample is 300 of 1,283 resolved domains, stepped across the pool rather than taken from the front of the alphabet — but it is not a random sample and should not be read as one.
• It generalises to recipients with a readable website, not to all 2,527 recipients. Roughly half the corpus could not be resolved to a live site at all, and companies that received an earnings-claims notice have an above-average rate of no longer trading.
• 60 of 300 sites could not be read, and they are excluded rather than counted as zero. 29 returned no text at all — JavaScript-rendered pages our reader cannot execute. Three declined to be read in robots.txt and were not read.
• Reading a homepage alone would have understated this badly. These companies put claims on product and science pages; the read follows up to four such pages per site.
• No company is named — not here, not in the underlying record we publish, not on request. The unit is the distribution. A named company beside this sentence would be an accusation, and it would be one we have no basis to make.

Why this is worth measuring

The Commission's theory of the notice is that putting an industry on written notice makes penalties available later. That theory has a measurable consequence: it should change what companies publish.

On this evidence, for most recipients, it has not — at least not in the part of their business a customer can see. Three in four still publish at least one claim with no evidence beside it.

That is not an accusation. It is the gap between a regulatory instrument and the pages it was meant to influence, and it is the first time anyone has measured its width.

---

Look your own company up. The FTC published five recipient lists as five separate PDFs. We made all of them searchable in one place — 2,527 companies, free, no account. Appearing on a list is not an indication that a company has done anything wrong.

And the question it raises next: what do your own published pages claim today, and what evidence sits beside each claim? We will read your pages and send back the inventory, free. It is a reading, not a verdict, and it is not legal advice. A reading is private to the company that requests it.