On October 21, 2024, the Federal Trade Commission’s Rule on the Use of Consumer Reviews and Testimonials, codified at 16 CFR Part 465, took effect. Fake reviews stopped being merely a deceptive trade practice addressed case by case and became a federal rule violation carrying civil penalties. It was, on paper, one of the most consequential consumer protection instruments of the decade: a direct answer to an estimated review-fraud economy that one industry compilation priced at $770.7 billion in consumer harm worldwide for 2025 alone (WiserNotify).
The rule just turned two. The birthday is a good moment to run the actual ledger, because the gap between the instrument and the problem is the single most important fact in product trust today.
Here is the complete record, drawn from the FTC’s own announcements and from DLA Piper’s July 2026 analysis of the commission’s enforcement approach:
- Ten warning letters, sent quietly on December 22, 2025 (Arnold & Porter)
- One administrative settlement: TruHeight, final order issued July 15, 2026, a $4 million judgment partially suspended upon payment of $750,000
- One pending federal lawsuit: Premium Home Service, filed with the Illinois Attorney General in May 2026, currently in motion practice
- Zero standalone actions focused primarily on reviews (the last one, Sitejabber, predates the rule)
Total cash actually collected under the rule in twenty-four months: $750,000. In the same period, Amazon proactively blocked more than 275 million suspected fake reviews, and that was just 2024 (Amazon).
This is not an argument that the FTC is failing. It is an argument about what law can do. A regulator samples; it cannot filter. And somewhere between those two functions sits every purchase decision made this holiday season, increasingly by machines reading the survivors.
What the Rule Actually Banned
The scope of 16 CFR Part 465 is broad and deliberately modern. Per the FTC’s announcement, it prohibits fake or false consumer reviews and testimonials, including reviews from people who do not exist, with AI-generated fakes named explicitly; buying positive or negative reviews, meaning any incentive conditioned on sentiment; undisclosed insider reviews; and review suppression. Violations carry civil penalties assessed per violation, a figure that inflation adjustment has pushed from $51,744 at adoption to $53,088 today.
Then-FTC Chair Lina Khan framed the rationale in terms that have aged well: “Fake reviews not only waste people’s time and money, but also pollute the marketplace and divert business away from honest competitors. By strengthening the FTC’s toolkit to fight deceptive advertising, the final rule will protect Americans from getting cheated, put businesses that unlawfully game the system on notice, and promote markets that are fair, honest, and competitive.”
Two sentences of that quote deserve a second read. Fake reviews “divert business away from honest competitors” is a competition claim, not just a consumer claim. And “put businesses on notice” is an accurate description of the mechanism: notice, then exemplary punishment of a sample. The rule was never designed to catch everyone. It was designed to make the expected cost of fraud nonzero. Whether it has even done that is the interesting question.
The Two Cases, Dissected
TruHeight is the rule’s first and only completed monetary enforcement. In April 2026 the FTC filed an administrative complaint against Nevada-based Vanilla Chip LLC, which does business as TruHeight, and its two co-founder principals, over supplements marketed to parents with claims like “The Only Supplement Clinically Proven to Help Height Growth.” The FTC’s announcement describes a review operation with three distinct supply channels: several thousand five-star reviews on the company’s own website that were “purportedly written by customers but were actually written by company employees” until at least November 2024; free and discounted products exchanged for five-star reviews on its site and on third-party platforms; and fake social media profiles “run by automated bots” posting “software-generated comments” on Facebook and Instagram.
Christopher Mufarrige, Director of the FTC’s Bureau of Consumer Protection, drew the connection explicitly: “TruHeight not only made unsubstantiated claims about its products’ capacity to boost height growth in children and teenagers but also amplified those claims with fake and incentivized reviews.”
The order imposes a $4 million judgment, suspended to $750,000 based on the respondents’ claimed inability to pay. Read the mechanics carefully, because they matter for deterrence math: the rule violation was not the case’s center of gravity. The health claims were. DLA Piper’s takeaway is that rule violations “will often appear as adjuncts to FTC cases primarily focused on other forms of deceptive conduct.” In TruHeight, the rule was the only legal basis for the monetary award, a damages hook bolted onto a broader deception case.
Premium Home Service is the pending one, and it may be the more instructive file. In May 2026 the FTC and the Illinois Attorney General announced a lawsuit, filed by the Department of Justice on the FTC’s behalf, alleging thousands of fake local business listings for home repair companies. The review allegation is the part worth studying: PHS allegedly directed employees, relatives, and search engine optimization contractors to post fake five-star reviews for one specific purpose, quoting the complaint’s language as relayed by DLA Piper, “to dilute the impact of real one-star reviews.”
That is not persuasion. It is statistical poisoning. The fraud does not need to convince anyone of anything; it needs to move an average. Hold that thought, because it is exactly the interface an AI shopping agent reads.
FTC Commissioner Mark Meador’s concurring statement focused on the competitive harm: “A particularly troubling aspect of this case is how the fraud as alleged defeated the efforts of diligent consumers who went to extra lengths to try to hire reputable, local providers with good reviews.”
The defendants filed a motion to dismiss on July 17, 2026, arguing that the conduct cited predates the rule’s effective date. Whatever the court decides, the argument itself is a reminder that the rule is strictly prospective: the multi-billion-review fraud economy built before October 2024 sits outside its monetary reach.
The Enforcement Math
Run the numbers the way a fraud broker would.
Independent estimates of the fake share of Amazon reviews range from 30 to 42 percent (Fakespot) to 7 to 11 percent (ReviewMeta), a range whose width is itself a trust problem (Review42). Amazon’s own proactive blocking, more than 275 million suspected fake reviews in 2024, with hundreds of millions more reported in its first Trustworthy Shopping Experience Report, published April 23, 2026, confirms the volume from the defense side.
Against that, the rule has produced two actions and ten letters in two years. At $53,088 per violation, the theoretical penalty exposure of a single mid-size review farm is astronomical; the actual expected penalty, given observed enforcement frequency, rounds to a rounding error. A seller pricing a fake review campaign as a customer acquisition channel faces near-certain upside and an observed downside of one $750,000 payment industry-wide in two years, extracted from a company that was primarily being punished for other conduct.
Deterrence theory has an answer to this, exemplary damages, and the rule nominally provides it. But the TruHeight suspension shows the ceiling: even when the commission wins $4 million, ability-to-pay realities collect $750,000. The sticker price of two years of federal fake review enforcement is less than the median US house.
The Volume Work Is Private
None of this means nothing is happening. It means the volume work is being done by platforms, privately, through litigation and filtering rather than penalties.
Amazon and the Better Business Bureau have now filed joint lawsuits against fake review brokers, including the operators of Skitsolutionbd.com, which sold “Five Star Amazon Verified Reviews,” fake negative reviews targeting competitors, fake seller feedback, and fake BBB business reviews, claimed access to thousands of reviewers worldwide, promised “100% safe” reviews, offered bulk discounts, and, remarkably, guaranteed replacements for any reviews that got removed. Read that last clause again: the counterfeit economy now ships with a warranty. Amazon has also sued Amzreview.ca over packages of fake verified reviews targeting its Canadian store.
Melanie McGovern of the International Association of Better Business Bureaus described the model: “By combining our capabilities and sharing intelligence, we are better equipped to identify and act against the people behind these deceptive practices.”
Private enforcement is real enforcement. But note what it optimizes for: the platform’s marketplace, the platform’s ranking integrity, the platform’s ad pricing. And note what it produces: removals that are invisible to everyone downstream. When Amazon blocks 275 million reviews, no consumer and no agent can tell which surviving corpus was filtered and how. The filter’s existence is announced; its output is unverifiable.
The Reader Changed in the Rule’s Second Year
Everything above would matter on its own. What makes it urgent now is that the rule’s implicit victim, a human reading reviews with their own eyes, is no longer the only reader.
GoBuy covered the machine-reader problem earlier this week: agents do not skim past fake praise the way a skeptical human might; they ingest aggregate signals, average rating, review count, sentiment ratio, verified-purchase flags, and compress them into a confident recommendation. Recall the Premium Home Service playbook: flood five-stars to dilute one-stars. That playbook was aimed at human sort-and-scroll behavior, but it is precisely engineered for the interface an agent reads. The fraud moved the aggregate; the aggregate moved the model; the model moved the purchase. Nobody in that chain saw a single fake review, and every one of them worked.
The rule, as structured, has nothing to say about this. Its remedies are retrospective and human-shaped: an investigation, a complaint, a judgment, eventually restitution to identifiable human purchasers. An agent-mediated purchase steered wrong by a poisoned sentiment ratio has no complaint form. The deception happened inside a context window nobody reads, downstream of a filter nobody can audit, upstream of a purchase the human delegated.
Regulation, in other words, sets the floor. It defines what counts as fraud and occasionally collects a sample. It cannot and will not compute, at query time, whether the specific review corpus in front of a specific agent is trustworthy. That is an engineering problem, and it belongs to the trust layer.
The Floor and the Ceiling
Two years in, the honest scorecard for 16 CFR Part 465 reads like this: it named the crime correctly, it armed regulators with a monetary hook that already works as an adjunct charge, and it has materially changed nothing about the information a shopper or a shopping agent encounters on a product page tonight.
Closing that distance requires exactly what platform filters and federal penalties do not provide: post-filter verification, computed on the evidence that survived, weighted by quality rather than volume, and delivered in a form machines can consume before they recommend. That is the design brief GoBuy was built to: fake reviews screened out and authentic ones weighted up before anything is scored, a Smart Score from 0 to 100 computed on that filtered evidence rather than on raw review counts, a curated top-7 result set instead of an infinite ranked scroll, and a GoBuy Verified badge reserved for products holding a score above 80 across 90 days. The same scores are exposed where agents already work, through the MCP server at gobuy.ai/api/mcp, so an agent can consult product trust the same way it consults a catalog, and the Chrome extension overlays the identical panel on Amazon pages so the human and the machine read the same filtered truth.
The FTC will keep sampling the fraud. Platforms will keep filtering it in private. The purchases, human and agentic, happen in the space between, this holiday season at meaningful agent scale for the first time. Trust in that space is not a legal status. It is a computation, and somebody has to run it before the buy button, not two years after.
If you are building shopping agents, the trust endpoint documentation is at gobuy.ai/agent-docs. If you are still reading reviews with your own eyes, start at gobuy.ai.