In September 2026 I read a forum thread where people were sizing up an "AI" helper that online stores use to fight chargebacks. One reply put it plainly: "It can't create proof that doesn't exist." That line gets at the heart of AI washing: making a product, process, or capability sound more intelligent, or more driven by AI, than it really is.
A chargeback is a customer disputing a card payment through their bank, which leaves the store to prove the sale was real. According to one reply, the AI part covers only one kind of dispute, orders marked as not received. Another says it assembles order information the store already has. A tool like that can still be useful. It is just narrower than the word "AI" suggests.
Ask one question about that tool, and about anything sold to you with an AI label: what does the AI part do that something simpler couldn't?
Why is AI washing so easy right now?
There is a knowledge gap. Technology has moved fast, and running a business leaves little time, or reason, to keep up with what these tools are and when to use them. AI is what everyone is talking about, so some vendors lean on that unfamiliarity. The label does the selling.
The cost lands on you. You can pay extra for a label, trust a tool with work it can't do, and hand your customer data to a company you didn't know was involved.
AI washing cases: what the SEC and FTC did
This isn't just theory. In March 2024 the SEC announced settled charges against two investment advisers "for making false and misleading statements about their purported use of artificial intelligence (AI)." They agreed to pay $400,000 in total civil penalties. Gary Gensler, SEC chair at the time, named it directly:
"Investment advisers should not mislead the public by saying they are using an AI model when they are not. Such AI washing hurts investors."
Six months later the FTC announced Operation AI Comply. It was five law enforcement actions against operations "that use AI hype or sell AI technology that can be used in deceptive and unfair ways." Lina M. Khan, FTC chair at the time, said: "The FTC's enforcement actions make clear that there is no AI exemption from the laws on the books."
AI washing examples: 4 common forms
1. Simple rules sold as AI
A new feature built on if/then logic, a script, or a calculation, and sold as AI. Picture a dispatch tool advertised as "AI scheduling" that sends each job to the nearest free technician. That is a rule. Useful, and still a rule.
2. Hidden human work
A system sold as AI that runs on its own, while people do a big part of the work behind the scenes. In January 2025 the SEC settled charges against a company that sold AI ordering for restaurant drive-thrus. The SEC's summary of its order says the company "falsely claimed that its own AI product eliminated the need for human order-taking. In fact, the vast majority of drive-thru orders placed through this version of [its product] required human intervention." The company neither admitted nor denied the findings. The SEC imposed no civil penalty, citing the company's cooperation and remedial efforts.
3. Old features with a new label
A feature that worked the same way last year, like search or filtering, now stamped "AI." Nothing changed but the name.
4. Borrowed AI
A company puts its own screen on top of someone else's AI and presents it as its own breakthrough. Using someone else's AI isn't deceptive in itself. The problem starts when the marketing makes it sound like something they built. The same summary says "[the company] failed to disclose that, for a period of time, the AI speech recognition technology in all units of [its product] that the company had then deployed was owned and operated by a third party."
Questions to ask an AI vendor, and why they aren't enough
After the one question above comes the usual list:
- What AI model are you using? (The model is the AI engine underneath.)
- Did you build it or license it?
- What data does it use?
- How much human work is involved?
- What are its limits?
- Will my data be used to train future models?
Those are fair questions. But what happens when you don't understand AI well enough to judge the answers? That is the real problem. If you don't know how AI differs from ordinary automation, you can ask every right question and still be talked into it by a good salesperson.
Does AI in the system mean AI is doing the work?
I've seen the opposite with one of my own clients. I built an automation for them, and at first they assumed it would include AI. When we looked at the process, I explained that it didn't need AI. Automation has been removing repetitive work and connecting systems for years. It doesn't need AI to do that.
I could easily have added some weak AI piece so I could say, "Look, it runs on AI now." Technically there would have been AI in the system. The automation would still do the heavy lifting. The AI would mostly be there for appearances.
AI being in a system doesn't mean AI is driving the value.
When AI is worth it, and when automation is enough
AI belongs where it does something traditional technology can't do as well, or where it clearly improves a specific task or bottleneck. That might be:
- Making sense of messy information, like emails or notes typed in free text.
- Writing or reworking content.
- Sorting, predicting, or spotting patterns.
- Handling variety that fixed rules struggle with.
Say an estimating tool adds your markup to a materials list. That is a formula. If it reads a customer's photos and drafts line items for you to check, it is doing something a formula can't.
Sometimes the right answer is simply an automation, or better process design, or better data. AI vs automation: which tasks to hand over walks through the difference task by task.
Don't use AI because a vendor says you should. Use it because you've found a problem where AI actually adds value.
What this post can't tell you
You can't see inside a vendor's product, and neither can I, so this post can't tell you whether any particular product is washed. From the outside, an "AI" feature could be a model, a set of rules, a person, or a mix, and only the vendor can show you which. Not every AI label is washed, either. Treat the label as a claim to check, not a verdict.
How to spot AI washing before you buy
You don't need to become an AI engineer. You need to know enough that nobody can sell you what you don't need because it says "AI" on the label. On your next sales call, ask the one question first. If the answer is vague, press on the two that are easiest to dodge: how much a person does behind the scenes, and whose AI it is and where your data goes.
An owner who understands their own systems can look at a proposal and say, "Wait. I don't need AI for this. This is a straightforward automation." Or, "Yes, this part benefits from AI. Now show me exactly what it's doing."
So the question shouldn't be, "Does this have AI?" It should be, "What problem is AI solving here that something simpler couldn't?" Once you can answer that, the hype gets a lot easier to see through.
Questions owners ask
Is AI washing illegal?
It can be. This isn't legal advice, but regulators have acted: the SEC settled charges over false AI claims in 2024 and 2025. The FTC's chair at the time said there is "no AI exemption from the laws on the books." If you think a vendor misled you, talk to a lawyer.
What does a good answer from a vendor sound like?
Here are two made-up answers. A clear one names the task: it reads the customer's voicemail, fills in the job type, and your office checks it. A vague one only repeats the label: our AI optimizes everything. If you only get the second kind, keep asking.
You might also like The MIT AI report and its 95% number, which shows what the widely quoted MIT 95% figure actually measured and how to check any AI claim.