Cox Media Group's own website once carried a line in bold: "Your Devices Are Listening to You." The company sold small business owners a product called Active Listening, billed as an AI that picked up purchase intent from smart phones, smart TVs, and speakers, then aimed local ads at those buyers. No device ever listened to a thing. What Cox sold was a consumer email list bought from a data broker, repackaged under a premium label, and marked up.

The Federal Trade Commission settled with Cox and two other firms for $880,000 over the past few months. The binding orders run for twenty years. Cox bought email lists from brokers, stamped them with the term "Active Listening," and pitched the whole package as artificial intelligence to business owners who had no way to check. The geographic data on those lists did not even match the zones their clients had paid to reach. The gap between the slide and the product is not a fraud story about one careless vendor. It is a pricing structure with a name.

"AI-powered tools give small businesses an edge." You have heard this line. Probably from a vendor who could not explain what the AI part actually did.

The Cox case is the thirteenth AI-washing action the FTC has filed since 2024. That number alone tells you something. Seven of the last eight cases target sales made to other businesses. Not to consumers clicking "buy now." To operators, people like you, signing contracts for tools meant to run their shops.

A May 2025 Gartner survey of 506 CIOs and technology leaders found that 72 percent of them were breaking even or losing money on their AI spending. Think about that. The people who run the largest tech budgets in the world cannot yet tell which AI tools pay for themselves. When the buyers with the deepest data and the biggest teams still struggle to measure return, the market opens a gap. Vendors fill that gap with a word. The word is "AI." And the label does the selling that the product cannot.

Small businesses spent an average of $2,340 on AI subscriptions in 2025. Roughly 31 percent of those tools went unused within 90 days. That is not a spending problem. It is a measurement problem.

Thirteen FTC cases did not happen because vendors are stupid. They happened because the market rewards the pattern.

The platitude does not fail because AI is useless. Some AI tools work well. The platitude fails because it skips the one question that matters: does the AI exist at all?

The problem is not the promise. The problem is that the label replaces the audit. Once "AI-powered" goes on the box, most buyers stop asking what sits inside the box. The vendor knows this. The pricing depends on it.

There is a name for this in the literature. Label arbitrage. Take a commodity product. Stick a premium term on it. Charge more. The term does the work that the product does not. Cox took a data broker's email list, a product you could buy for a fraction of the price on the open market, called it AI, and sold it at a markup to business owners who trusted the label. The mechanism is not new. The label is.

The better rule is plain: audit the label before you pay the price. Once that is clear, three moves follow from it.

Before money changes hands, ask the vendor to name the data that goes into the AI system and the result that comes out. Not the pitch deck version. The actual data flow. A real tool can describe its inputs and outputs in two sentences. "We ingest your CRM contact list and return a ranked lead score based on purchase recency." That is a real answer. A resold email list cannot produce one.

This is the move that costs something, because it means sitting across from a sales rep and saying the quiet part out loud: I do not trust the label, and I need you to show me what is behind it.

If the vendor says "proprietary data," ask who collects it and how. Cox's data came from third-party brokers. The sales team never mentioned that. They told prospects the data came from device microphones, from "every casual conversation" picked up by smart speakers. One direct question about the source of the data, asked before the contract arrives, splits real tools from relabeled ones in minutes. The vendors who built something real will welcome the question. The ones who relabeled something will redirect you to the slide.

Do not sign an annual deal on a demo. Buy one month. Pick one metric you can track in your own books without the vendor's dashboard. Revenue from a specific campaign. Leads from a named source. Calls booked in a set time window. If the tool works, the metric moves. If the metric holds still, the tool did not work. The vendor's own report is not proof. Your numbers are proof.

Running these three moves for 90 days does something the platitude never did:

You see which vendors answer the hard questions and which ones steer you back to the pitch deck. You see which tools move a number you can check in your own records, not just on their screen. You see the size of the gap between what got sold and what got shipped. And you stop paying for a word that replaced the product it was supposed to describe.

At the end of 90 days, ask three things.

→ Which tool moved a number I can find in my own books?

→ Which tool looked like progress but left no trace in revenue or pipeline?

→ Which vendor could not answer the input-output question?

That is the difference between advice that sounds right and a system that proves itself.

The pitch deck has not changed. The slide still reads the same six words. What shifted is what you know to ask before you sign.

I have sat through that pitch. Most operators in this position have sat through that pitch. The feeling that something was off, that the promises were too smooth and the proof was too thin, was not a lack of knowledge. It was good judgment waiting for the right question.

The edge was never in the label. It was in knowing what the label replaced.