A company can make an ordinary feature sound groundbreaking simply by putting “AI-powered” in its description. When technology feels like a selling point, companies may feel pressure to use that label in marketing. But broad language can leave buyers unsure what a product actually does.
Substantiation means objective evidence supporting an advertising claim when the claim is made. For your business, that evidence helps separate real capabilities from polished language. The same distinction matters when you assess a vendor or describe your own products.
This guide explains practical ways to check AI claims and outlines the US business risks of getting them wrong. Treat an AI label as a claim to verify, not as proof of a product’s capability.
Key Takeaways
- An AI label should match a feature you can identify.
- Marketing claims need evidence that supports their meaning.
- Human operators may still play a major role.
- Investors and consumers may rely on AI claims.
- Keep test records that back up performance promises.
- Distinguish a current feature from a planned one.
What Is AI Washing, and What Does It Look Like?
AI washing is misleading, unsubstantiated, or false language about artificial intelligence in a product, service, or business communication. It can appear in marketing, investor statements, and product descriptions.
The claim should match what the technology does today. The Federal Trade Commission applies the same truthfulness, nonmisleadingness, and substantiation standards to AI marketing claims as to other advertising.
Examples of AI washing include overstating a product’s AI role or capabilities. A company might claim a proprietary machine-learning model powers its service when human operators actually perform the work.
Do not describe work as automated when people perform the task the claim says AI handles.
That distinction matters in products that combine software and people. Human review can be part of an AI service, but the marketing should say what the system does and what staff still do.
Overstating a product’s AI role or capability
A product may use artificial intelligence for one small feature while relying on ordinary software for most tasks. A claim that implies AI runs the whole product can give buyers the wrong impression.
Specific wording gives a clearer picture. A company might say an AI feature sorts support requests, rather than claiming its full service runs itself.
Hiding human work behind an AI claim
Some services rely on human operators to review, label, or create work that customers think software completes. The marketing can mislead when it presents that labor as automated output.
Human involvement is not automatically a problem. The issue is whether the claim hides the role people play or exaggerates the product’s capabilities.
Rebranding basic or unreleased technology
A company may rename a basic rule-based tool as artificial intelligence without explaining what it actually does. It may also promote a feature as available before customers can use it.
The term resembles greenwashing, which creates a misleading impression about a product’s environmental benefits. In both cases, the marketing can say more than the underlying product supports.
Some appliances, for instance, use AI features without guaranteeing better results for every household. A washer’s AI feature matters when it solves a real laundry need, not just because the label sounds advanced. A specific, limited claim can be accurate even when the product does not rely on AI for most of its work.
Why Does AI Washing Matter to US Businesses?
A misleading AI claim can create more than disappointed customers, because competitors, investors, and regulators may also scrutinize it. A claim can affect buying decisions, trust, and legal risk at the same time.
In a market where companies use artificial intelligence to stand out, the impact can reach people who never use the product themselves.
- Buyers may choose products that fail to meet their operational needs.
- Investors may misjudge a company’s capabilities or business prospects.
- Consumers may lose trust in a company and in legitimate AI products.
- Competitors may bring false-advertising claims under the federal Lanham Act.
A competitor may argue that false claims gave a company an unfair edge. Separate from lawsuits, the National Advertising Division reviews national advertising claims and issues decisions or recommendations.
Those recommendations are not law, but they can create a public record that regulators and plaintiffs’ lawyers may reference. The National Advertising Division’s review process also allows referrals to the FTC when an advertiser refuses to take part or follow recommendations.
State rules can add another layer. For example, New York’s law on synthetic performers in ads is scheduled to take effect June 9, 2026, and requires conspicuous disclosure in covered cases. The New York disclosure law for synthetic performers addresses a specific form of AI-related advertising, not every AI claim.
Layoff announcements need the same care as product claims. A company saying AI drove workforce cuts does not establish that AI caused them; look for deployed technology and actual workflow changes.
That distinction matters in 2026, when an AI explanation may shape how employees, investors, and customers judge a company’s choices. Treat an AI-led layoff explanation as a claim about business operations that needs evidence, not as proof of what caused the cuts.
How Can Businesses Check AI Claims?
Ask the vendor to name the exact feature behind an AI claim before discussing broad promises about the product. A clear feature gives you something concrete to test.
Check what the feature does, who does the work, and what evidence supports its claims. This applies to technology you buy and to claims your company plans to publish.
Use these five checks before relying on a claim or putting it in marketing:
- Identify the exact feature described in the claim.
- Ask what role AI plays in the core product.
- Clarify what human operators do and what data the feature uses.
- Request testing, validation, and performance evidence.
- Check whether demonstrations and productivity promises match current use.
A product can include human review without being AI-washed. The claim should explain what the system automates and where human operators step in.
For example, test a legal tool on work your team actually handles. A lawyer’s AI workflow may include research or contract review, while a lawyer still checks the work and makes final decisions.
Verify the feature and its role
Confirm that customers can access the feature now, rather than only seeing it in a demo or roadmap. A demonstration should show the real product and its current limits.
Ask which data the feature uses and whether that data enters a model. A clear answer helps you compare the claim with the product’s actual design.
Match performance promises to evidence
Performance figures need context, including the task, test conditions, and product version. A headline score alone does not show how a tool will perform in your work.
A benchmark for AI text tools, for example, is useful only when you know the sample, detectors, and writing types tested. Results under one test setup do not guarantee the same result elsewhere.
| Claim type | Question to ask | Evidence to review | Warning sign |
|---|---|---|---|
| Feature availability | Can customers use it now? | Current product access | Roadmap shown as live |
| AI’s role | What task does AI perform? | Workflow and system details | Vague “AI-powered” wording |
| Demonstrated performance | Does it work in normal use? | Test results and limits | Only a perfect demo |
| Productivity claims | How was time saved measured? | Task and timing records | No objective support |
Companies may face scrutiny for promoting unreleased features, demos that imply unreliable performance, or productivity claims without objective substantiation. Keep the supporting test and performance records that were available when the claim was made.
What Can US Regulators and Courts Do?
US enforcement already includes an SEC settlement over investment advisers’ statements about their AI use. Different legal routes apply to advertising, investment advice, securities statements, and state consumer protection.
The FTC applies ordinary advertising standards to AI claims, including business-to-business marketing. It may also pursue vendors that supply deceptive marketing materials under the means and instrumentalities doctrine.
The FTC launched Operation AI Comply in September 2024. In 2025, it initiated more than a dozen cases tied to AI washing, targeting companies accused of exaggerating or fabricating AI capabilities.
| Route | Who brings it | Claim focus | Possible significance |
|---|---|---|---|
| FTC advertising enforcement | Federal Trade Commission | Deceptive consumer or business marketing | Agency action and penalties |
| SEC adviser enforcement | Securities and Exchange Commission | Adviser and investor communications | Orders and civil penalties |
| Private securities claims | Investors or class members | Misstatements under Rule 10b-5 or Section 11 | Damages or case dismissal |
| State consumer enforcement | State attorneys general | Deceptive claims under Little FTC Acts | State investigation or action |
On March 18, 2024, the SEC announced settled charges against Delphia (USA) Inc. and Global Predictions Inc. The SEC ordered Delphia to pay $225,000 and Global Predictions to pay $175,000 in civil penalties.
The SEC charged both firms under Sections 206(2) and 206(4) of the Investment Advisers Act of 1940, the Marketing Rule, and the Compliance Rule. Delphia had claimed since August 2019 that its investment-advisory algorithms used AI, including machine learning, and incorporated client data.
In July 2021, Delphia told SEC staff it had not used client data and lacked an algorithm to use it. The firm continued saying client data informed its investment algorithms through August 2023, although the SEC found the data had never been an input.
The SEC settlement also covered an unsupported claim that Delphia used client data to train AI models for investment decisions. Global Predictions called itself the “first regulated AI financial advisor,” which the SEC determined was false.
The SEC’s 2024 adviser settlements show why claimed AI use should match documented practice.
Specific allegations matter in securities cases; broad claims about a product’s AI line may not show what feature failed.
A federal court dismissed a SaaS securities class action because the challenged statements were forward-looking with meaningful cautionary language, puffery, or too generic to establish falsity. Allegations that its AI product line missed touted capabilities lacked detail because they named no specific feature or failure.
Private claims under Rule 10b-5 or Section 11 may arise from AI-related misstatements, including for companies nearing an IPO or SPAC deal. State attorneys general can also investigate deceptive AI claims under state Little FTC Acts.
Public statements about AI capabilities, revenue, or market share need factual support or clear forward-looking language with meaningful cautionary wording.
Conclusion
Clear, specific claims make it easier to distinguish a real capability from AI hype. A convincing label alone does not show that a product works as described or helps with a real need.
Choose one prominent AI claim in a product description, pitch deck, or campaign. Confirm the feature, the role of human operators, and the records that support the claim.
Review one current AI claim against the feature it describes and the evidence that supports it.
FAQ
Are there AI layoffs expected in 2026?
A layoff announcement alone does not show that AI caused the cuts. Companies may name AI as a reason, but that statement is not the same as proof. Look for evidence that the technology was deployed and changed actual workflows. The timing, the affected roles, and the work shifted to software can help explain what happened.
Can human-reviewed software still use AI?
Yes, human-reviewed software can still use AI. People may check, edit, or approve output from an AI system, and that review can be part of the product’s normal workflow. The claim should say what the software handles and what human operators still do. Calling the whole service automated may mislead when people complete key tasks.
Is a simple algorithm enough to call a product AI-powered?
A simple algorithm does not prove that an AI-powered label is accurate. The company should describe the function the algorithm performs, without implying advanced learning or abilities it lacks. A rule-based tool may sort records or follow set steps, for example. The wording should help customers understand that function rather than suggest a more capable system.
Why is AI washing compared with greenwashing?
AI washing is compared with greenwashing because both use marketing to create an impression the product or practice does not support. Greenwashing can overstate environmental benefits, while AI washing can overstate a product’s AI use or ability. In either case, the claim can shape a buyer’s choice. Clear details help people judge what the product actually offers.
Can a company describe a feature as AI-powered before it launches?
Yes, a company can describe an unreleased feature as planned if its wording makes the future status clear. It should not imply that customers can use the feature now when it has not launched. A roadmap or future product announcement should distinguish planned work from current capabilities. That difference helps customers avoid relying on a feature that is not yet available.
Are AI claims in investor communications treated like product marketing claims?
Investor communications can raise securities-law concerns, even when they are not product ads. Public statements about AI capabilities, revenue, or market share should be accurate or clearly marked as forward-looking. Forward-looking claims need meaningful cautionary language. The details matter because investors may rely on those statements when assessing a company’s business, growth plans, or potential value.
