Read these three sentences the way a regulator would:

Each one is a factual claim. Each one implies the business has evidence behind it. And each one closely tracks the language in an actual Federal Trade Commission case. “AI washing,” meaning overstating what your AI does or whether you use AI at all, has been one of the steadiest areas of FTC enforcement through two administrations. It has also reached small sellers and business-to-business vendors, not only large consumer brands.

This piece explains the rule, walks through the cases, and ends with a way to review your own marketing before someone else does.

The rule fits in one paragraph

Section 5 of the FTC Act prohibits “unfair or deceptive acts or practices in or affecting commerce.” Since 1984, the FTC's Policy Statement Regarding Advertising Substantiation has said that advertisers must have “a reasonable basis for advertising claims before they are disseminated.” If you state a level of support (“tested,” “proven,” “98% accurate”), you need at least that level of support. If your ad implies support, you need whatever support the ad actually communicates to the reader.

Nothing about AI changes that. When the FTC launched its AI sweep in 2024, then-Chair Lina Khan said the actions made clear “there is no AI exemption” from existing law. The current Commission, under Chairman Andrew Ferguson, has kept bringing these cases.

What the cases actually look like

The FTC's AI-washing cases fall into a few recognizable patterns. Here are the ones that matter most for a small business that sells something with AI in it.

Pattern 1: “It replaces a professional”

In September 2024 the FTC announced Operation AI Comply, five actions in one release. The one that drew the most attention was against DoNotPay, which marketed itself as “the world's first robot lawyer.” According to the FTC, the company had not tested whether its chatbot performed at the level of a human lawyer and did not employ attorneys. The final order, approved 5-0 in January 2025 and announced February 11, 2025, required DoNotPay to pay $193,000, notify past subscribers, and stop claiming its service performs like a lawyer without sufficient evidence.

The lesson isn't limited to legal services. Any claim that your AI does what a licensed professional, a trained employee, or a human team does is a performance comparison, and you need evidence for it.

Pattern 2: A specific accuracy number

In April 2025 the FTC charged Workado, formerly Content at Scale AI, over its AI Content Detector, which it advertised as “98 percent” accurate at telling AI-written from human-written text. The FTC said independent testing showed about 53 percent accuracy on general-purpose content, and that the model had been trained or fine-tuned mainly on academic writing. The order, which the Commission later finalized on a 3-0 vote, bars effectiveness claims unless they are backed by “competent and reliable evidence” at the time they are made, and requires the company to keep that evidence.

A precise number reads as a tested result. If you can't produce the test, don't publish the number.

Pattern 3: AI as a business opportunity or earnings engine

Three of the five Operation AI Comply cases (Ascend Ecom, Ecommerce Empire Builders, and FBA Machine) involved claims that AI-powered online stores would produce passive income. The FTC alleged those schemes cost consumers tens of millions of dollars. In August 2025 the FTC sued Air AI and related companies in federal court in Arizona. It alleged they marketed “conversational AI” as able to replace human customer-service staff and told small-business buyers they would quickly earn back their investment, while rarely honoring promised refunds. Those are allegations in a complaint, not findings by a court. In January 2026 the FTC announced a settlement with the Growth Cave defendants. It includes a permanent ban on selling business opportunities and a $48.6 million judgment, partially suspended based on inability to pay, along with a prohibition on misrepresentations about the use of artificial intelligence.

When “AI” sits next to an earnings or results claim, you need evidence for both halves of the sentence.

Pattern 4: The AI doesn't exist, or doesn't do what you say

The most recent example is also the most useful for business-to-business sellers. In May 2026 the FTC charged CMG Media Corporation (Cox Media Group) and two smaller marketing firms, MindSift and 1010 Digital Works, over an ad-targeting product called “Active Listening.” According to the FTC's complaint announcement, the companies told advertising customers the AI-powered service picked up conversations from smart devices to target local ads, and that consumers had opted in. The FTC alleged the product used no voice data at all and instead resold email lists bought from data brokers at a markup. The two smaller firms were charged with providing CMG the “means and instrumentalities” to mislead its customers through sales materials. The orders were finalized on August 27, 2026, with payments totaling $930,000 ($880,000 from CMG and $25,000 from each of the others) to be used for customer redress. Each company is barred from misrepresenting its services' features, its use of voice data, and its geographic targeting.

The firms resolved the case through consent orders. As with any FTC settlement, the facts recited are the agency's allegations.

Selling to businesses is not a safe harbor

A lot of owners assume consumer-protection law is about consumers. The FTC's own cases say otherwise. CMG's customers were advertisers. Air AI's alleged victims were small-business buyers. The theory is the same: a deceptive claim that induces a purchase is deceptive whether the buyer is a household or a dental practice.

The CMG case also shows a second exposure for small vendors. If you supply sales decks, scripts, or white-label materials that a larger partner uses to make AI claims to its customers, you can be named for providing the means to deceive, even if you never spoke to the end customer.

State attorneys general use the same theory under state deceptive-practices laws. In September 2024 the Texas Attorney General announced a settlement with Pieces Technologies, a Dallas health-AI company that had marketed a “severe hallucination rate” of “<1 per 100,000” to hospitals. The AG's office said the metrics were likely inaccurate. Pieces denied wrongdoing and publicly disputed the AG's description of the agreement, which carried no financial penalty but required clearer disclosures about accuracy and appropriate use. The customers were hospitals, not individual consumers.

Where the line is: the Rytr reversal

It would be a mistake to read all of this as “any AI marketing is risky.” The current Commission has been explicit that it doesn't want to punish AI tools for what users might do with them.

The clearest signal is Rytr. In 2024 the FTC, on a 3-2 vote, entered an order against Rytr, an AI writing tool with a “Testimonial & Review” feature, on the theory that it gave users the means to post fake reviews. Commissioners Ferguson and Holyoak dissented. In December 2025 the Commission reopened and set aside that order entirely. It found that the complaint's facts did “not support a finding that Rytr violated Section 5,” and that “Rushing in to impose aggressive law enforcement unsupported by facts or law is improper and is not in the public interest.” The order cited the July 2025 federal AI Action Plan, which asked the FTC to revisit orders that unduly burden AI innovation.

Put Rytr next to CMG and Workado and the line is clear. The FTC is not pursuing AI because it is AI. It is pursuing false and unsupported claims about what a product does, whether a product uses AI at all, and what results a buyer will get. Those are claims you control.

A red-pen review of your own copy

Here is how a lawyer would mark up the three sentences from the top of this article, and what a defensible version might look like.

“Our AI-powered platform automates your entire lead pipeline.”

“Proprietary AI catches 98% of errors before they reach your customers.”

“Clients using our AI system typically see their revenue double in 90 days.”

Before you put “AI” on a page: the claim file

For each AI-related claim on your website, sales deck, or proposal template, build a one-page file and keep it:

  1. Does the product actually use AI, as a reasonable buyer would understand that word? Rules-based automation with an “AI” label is the simplest case for an enforcer to bring. The FTC's 2023 business guidance asked this exact question.
  2. What exactly is the claim, express and implied? Write it down in plain terms. “Replaces your front desk” implies a performance level.
  3. What evidence did you have before publishing? Tests, data, dates, conditions. If the claim is comparative (“better than manual review”), the evidence has to be comparative.
  4. Who else repeats it? Resellers, affiliates, and partners using your materials. Their claims can become your problem, and yours can become theirs.
  5. When will you re-check it? Models change. A claim that was true for last year's version may not be true now.

If you also run a customer-facing chatbot, disclosure rules are a separate issue; our guide to chatbot disclosure laws covers them. If staff are writing marketing copy with AI tools, put review requirements in your employee AI use policy so an AI-drafted claim gets the same scrutiny as any other.

The bottom line

The FTC's AI cases are not about technology. They are about the oldest rule in advertising law: say what's true, and have proof before you say it. Under both the prior and current leadership, the Commission has treated “AI-powered” as a claim like “clinically proven” or “guaranteed results.” Write your copy so you could hand the evidence file to an investigator the day after it goes live.

This article reflects FTC and state enforcement actions announced through October 2026. FTC complaints and consent orders describe the agency's allegations; settlements are not findings of liability. This is general information, not legal advice for your business.

This article is for educational purposes only and does not constitute legal, tax, or financial advice. Consult a qualified professional for guidance specific to your situation.