AI + modern marketing
An Offer You Can’t Refuse? Let’s Read the Fine Print.
Five AI marketing red flags, checked against FTC cases, and the questions business owners should ask before buying another automation promise.
Watch: An Offer You Can't Refuse: The Truth About AI Promises ↗
Apparently, we can all automate the entire business, collect a thousand leads, and head to the beach. Somewhere between the first coffee and lunch, another stranger has figured out how to make us rich with one prompt. Convenient.
In my recent video, I asked you to send me the most ridiculous AI or marketing promise you’ve seen. Because behind the laugh is something I take seriously: capable business owners are being sold relief from overwhelm, and sometimes the thing they buy gives them more of it.
I understand the attraction. You’re running the business. The platforms keep changing. Someone arrives with a polished presentation and tells you they can take the whole mess off your hands. Of course you want to hear the offer.
I use AI. I see what it can make possible. I also want to know what happens after the demo, when your actual customers, your actual information, and your actual reputation enter the room.
So I went to the enforcement record. The FTC launched Operation AI Comply in September 2024 with five actions involving deceptive AI claims and schemes. The cases below give us a practical way to examine the next sales pitch. They have different histories and outcomes, so let’s keep those details attached to the story.
1. The Ghost AI: what is actually doing the work?
You’re being sold a capability. Before you buy it, someone should be able to explain what it does.
In August 2026, the FTC finalized settlements with Cox Media Group and two other firms over an advertising service called Active Listening. The companies had claimed they could target ads using conversations captured from smart devices, with consumers’ consent. According to the FTC, the service wasn’t based on voice data and consumers hadn’t opted in. The three companies were required to pay a combined $930,000.
That is a very different product from the one in the presentation.
Ask for a demonstration using a realistic example from your business. Which steps are automated? Where does a person step in? Where does the customer data come from, and what permission covers its use? A legitimate vendor may use third-party models or human review. Those facts should be clear. Refusing to disclose an entire training dataset is not, by itself, proof of deception.
2. The Fantasy ROI: whose income are we talking about?
A revenue screenshot can be impressive. It can also leave out the advertising spend, refunds, software bills, labor, and all the customers who got a very different result.
The FTC alleged that Ascend Ecom used AI-powered storefront promises in a scheme that cost consumers at least $25 million. Its June 2025 settlement announcement described a proposed $25 million judgment, partially suspended based on inability to pay. A judgment amount is not the same as money returned to customers.
In March 2026, the FTC announced a proposed settlement that would ban Air AI and its operators from selling or marketing business opportunities. The allegations concerned earnings claims, refund guarantees, and other representations. That is a specific restriction, not a blanket ban on every kind of marketing.
Before you buy a projected result, ask what a typical customer actually achieved, over what period, after which costs. Ask how many customers were measured and how many failed. Then read the refund conditions before you need them.
3. The Magic Percentage: accurate at what?
A percentage has a lovely way of making a sales slide feel scientific. Put 98% beside a product and suddenly the room relaxes. I’d like to see the test.
Workado advertised its AI content detector as 98% accurate. According to the FTC’s April 2025 complaint, independent testing on general-purpose content found 53% accuracy. The FTC alleged the model had been trained or fine-tuned to classify academic content effectively, despite broader marketing claims. The final order, approved in August 2025, requires reliable evidence for accuracy and efficacy claims.
The useful question goes beyond whether a test exists. Does it resemble the work you need done? A tool tested on one kind of writing may behave differently on your customer emails, product descriptions, or industry terminology. Ask about false positives, missed cases, and who checks the output before it affects a customer.
4. The Expert Replacement: who owns the mistake?
Replace your lawyer. Replace your sales team. Replace your agency. There’s usually a very tidy price beside a very untidy assumption about what those people actually do.
The FTC’s DoNotPay case concerned claims that its chatbot could substitute for a human lawyer. The FTC said the company hadn’t tested whether its service performed at that level or retained attorneys to test its legal features. The finalized order required $193,000 in monetary relief and evidence to support claims that the service performs like a real lawyer.
For your business, map the responsibility before you hand over the task. Who approves a customer promise? Who handles an exception? Who can stop an agent that is doing the wrong thing very efficiently? A tool can help someone work. Deciding who remains accountable is still a management decision.
5. The Manufactured Reputation: did the experience happen?
A customer review is supposed to describe an experience someone had. A tool’s ability to write a convincing experience doesn’t make it real.
Rytr needs a careful update here. The FTC’s 2024 case concerned a tool that could generate detailed reviews with information unrelated to the user’s input. But in December 2025, the Commission reopened and set aside that order, saying the complaint failed to support a Section 5 violation and the order unduly burdened AI innovation. Presenting that old restriction as current would be misleading.
The separate FTC Consumer Reviews and Testimonials Rule still addresses fake reviews, including reviews attributed to people who don’t exist or who never had the claimed experience. The Rytr reversal is not permission to invent satisfied customers.
Use technology to organize real feedback, identify recurring questions, and make it easier to invite an honest review. Keep the customer’s experience intact. Your reputation deserves better than a fictional supporting cast.
Before the next subscription, look at the foundation
If you’ve bought something that sounded better in the webinar, there’s no shame in that. The pressure to keep up is real. You can still stop, look at what you have, and make a better next decision.
Write down one process you want to improve. Follow it from the first customer action to the final handoff. Check the information it uses. Name the person responsible. Then test a tool on a small, reversible piece of that work and measure what changes.
That is how I want to approach modernization: understand the business, organize the moving parts, and give the technology a clear job. Sometimes you need a rebuild. Sometimes the good pieces are already there and need someone to connect them.
The next time an offer sounds impossible to refuse, give yourself permission to ask a few inconvenient questions. A useful product should survive them.
And keep sending me those AI promises in the comments on the video. We can have a laugh, read the fine print, and work out what is actually worth your time.