AI washing is the practice of using AI-related terms—such as 'AI-powered', 'machine learning-driven', or 'intelligent'—in marketing without substantiating what the AI actually does, how it was tested, or what evidence supports the claimed result. Regulators including the FTC (Operation AI Comply, September 2024) and the SEC (Delphia and Global Predictions enforcement, March 2024) have taken action against companies whose AI claims were not backed by the evidence the wording implied.

What makes an AI claim 'high risk' for buyers?

An AI claim carries a higher evidence burden when it uses absolute or comparative language (zero, perfect, always, highest), professional-equivalence wording (like a lawyer, as good as a doctor), compliance certification language (HIPAA-compliant, GDPR-ready, fully ADA compliant), or quantified outcomes (saves 40% of costs, 99% accuracy) without naming the test conditions, populations, and failure modes that define the figure.

How is AI washing different from ordinary marketing exaggeration?

Ordinary puffery—words like 'leading', 'innovative', or 'cutting-edge'—is generally not actionable because it is understood as subjective opinion. AI washing claims involving measurable accuracy, safety outcomes, professional equivalence, or legal compliance are held to a higher standard: the FTC requires competent and reliable evidence to exist before the claim is published. Specific AI performance or compliance claims are not treated as puffery.

How do I spot AI washing when evaluating a vendor?

Look for five patterns in the marketing copy. Accuracy numbers without a named test set or error rate ('98% accurate'). Compliance labels without the auditor, scope, or what the certification does not cover ('HIPAA-compliant'). Professional-replacement phrasing without task limits or a human-review boundary ('like a lawyer'). ROI or cost-saving claims without a baseline, sample size, or industry scope ('saves 40% of costs'). And 'AI-powered' or 'AI-native' language without naming which workflow step the model runs, what input it takes, or what a person still reviews. Each pattern means the vendor should be able to provide specific evidence when asked—it does not mean the product cannot deliver.

What are common AI washing red flags in vendor marketing copy?

The most common patterns from FTC and ASA enforcement records: specific accuracy figures without a visible benchmark or false-positive rate; automated compliance wording that does not name what automation cannot cover; ROI guarantees without a comparison baseline or sample size; 'first,' 'only,' or 'best' claims without naming the comparison set or date; and AI product descriptions that list features rather than tested outcomes with stated limits. These are the wording patterns mapped in the signal catalog on the Methodology page.

Which companies have faced regulatory action over AI washing claims?

FTC enforcement cases with AI claim findings include Content at Scale AI (AI detector accuracy, 2025), accessiBe (automated accessibility compliance, 2025), DoNotPay (AI lawyer claims, 2025), Workado (AI content detector, 2024), Evolv Technology (AI security screening accuracy), and IntelliVision (bias-free facial recognition claims). The SEC took action against Delphia and Global Predictions in March 2024 over AI investment capability claims. ASA in the UK ruled against PixVideo AI over AI video advertising in 2024. These are official-source enforcement examples, not a company ranking or judgment.

What does an AI washing claim pattern look like in a real case?

One documented pattern: Content at Scale AI marketed an AI content detector with a specific accuracy figure (FTC case, 2025). The evidence gap was that the stated rate did not publicly name the content types, model versions, or mixed-authorship cases covered, and did not disclose a false-positive rate. Claim type: Accuracy / Performance. Wording signal: detector accuracy stated as a percentage without benchmark scope. Buyer question to ask: 'What exact content types and model versions were included in the test that produced this number, and what is the false-positive rate for each?'

How do I check an AI vendor's claims before relying on them?

For each AI claim on the vendor's public page: identify the exact words that raise the evidence burden (percent figures, compliance labels, automation language, or superlatives); categorize the claim type (accuracy, compliance, automation, ROI, or vague AI-powered); name the evidence that would support the claim in your specific use case; and write one concrete question to ask the vendor. Pasting the vendor's public page URL or the claim text into the checker on this site steps through this process and outputs a Claim Receipt you can use in vendor conversations or add to procurement notes.

AI washing examples: claim patterns and evidence gaps

Last reviewed May 24, 2026

AI washing examples are useful only when they point to a specific claim pattern and the evidence needed to support it. This page uses official and high-evidence sources to show the wording pattern, not to compare companies.

Check the AI product claim you found How the evidence method works

Or load a documented sample into the checker

Fastest path: copy one exact vendor sentence that matches this pattern, then open the checker. Add the public URL only if you want readable page context recorded alongside the wording. The result is an evidence-burden note you can reuse in vendor follow-up or internal review, not a verdict. Not sure what a result looks like? See a sample receipt.

What to verify before you rely on the claim

Exact claim text from the public page or official source.
Claim type and the wording that raises the evidence burden.
Specific evidence that would support the claim in the same use case.

Sources behind AI washing examples

FTC Content at Scale AI case page
· August 28, 2025
Detector accuracy claim pattern.
FTC accessiBe case page
· April 22, 2025
Automated accessibility compliance claim pattern.
FTC DoNotPay case page
· February 11, 2025
Professional-replacement claim pattern.
SEC March 18, 2024 press release
· March 18, 2024
AI use, forecasting, and first-of-kind claim patterns.
SEC Delphia administrative order
· March 18, 2024
Source for client-data, predictive-investing, and AI-capability claim wording.
ASA AI as a marketing term report
· November 13, 2024
Official report context for vague AI marketing language.

Documented AI washing examples examples

"AI content detector can identify AI-generated text"

Accuracy / Performance

Source and date: FTC Content at Scale AI · August 28, 2025
Evidence signal: Detector capability wording without enough visible test context.
Evidence gap: Benchmark scope, content categories, model coverage, and error rates.
Buyer question: For the AI content detector claim, what exact content types and model versions were tested?

Load this sample in the checker

"AI-powered web overlay for accessibility compliance"

Compliance / Safety

Source and date: FTC accessiBe · April 22, 2025
Evidence signal: Broad automated compliance result.
Evidence gap: Standard version, audit method, manual remediation boundary, and ongoing maintenance.
Buyer question: For the any website compliant claim, what remains outside automation and requires human review?

Load this sample in the checker

"AI lawyer handles legal matters without professional review"

Automation / Replacement

Source and date: FTC DoNotPay · February 11, 2025
Evidence signal: Professional-replacement language.
Evidence gap: Task scope, quality testing, professional review, and limits on legal-document use.
Buyer question: For the robot lawyer claim, which tasks were tested and which require professional review?

Load this sample in the checker

"turns your data into an unfair investing advantage"

Accuracy / Performance

Source and date: SEC Delphia administrative order · March 18, 2024
Evidence signal: AI investment-advantage wording tied to client data and predictive capability.
Evidence gap: Data actually used, model capability at the claim date, decision boundary, and disclosure updates.
Buyer question: For the unfair investing advantage claim, what client data and model capability existed when the wording was live?

Load this sample in the checker

Evidence map for AI washing examples

Claim pattern	Evidence needed	Buyer question
Accuracy / Performance	Benchmark, test population, model list, sample size, and error rates.	Does the evidence match the exact content or workflow where we would rely on the claim?
Automation / Replacement	Human review boundary, failure handling, task limits, and escalation path.	Which steps are automated, and which steps still require qualified review?
Compliance / Safety	Applicable standard, audit method, known exclusions, and maintenance responsibilities.	Which standard is named, and what evidence shows coverage against that standard?
First / Only / Best	Comparison universe, time frame, source of comparison, and update process.	What happens if a competitor changes the comparison tomorrow?
Vague AI-powered	Specific model role, input, output, human step, and user-visible benefit.	If the word AI is removed, what concrete product capability remains?

Evidence buyers need for AI washing examples

Exact claim text from the public page or official source.
Claim type and the wording that raises the evidence burden.
Specific evidence that would support the claim in the same use case.
A buyer question that mentions the actual claim, not a generic AI question.

Buyer questions for AI washing examples

Which claim type is this: accuracy, automation, compliance, ROI, first/only/best, or vague AI-powered?
What exact word raises the evidence burden: percent, any, first, fully automated, or AI-powered?
What document, benchmark, audit, or disclosure would support this claim at the date it was published?
How could the claim be rewritten with scope, date, and human review boundary included?

Safer wording for AI washing examples

Reported X% on a named benchmark, with limitations stated.
Automates specified low-risk steps and routes exceptions to human review.
Supports compliance work by identifying selected issues; final review and maintenance remain separate.
Uses an AI model in named workflow steps instead of claiming broad AI transformation.

AI washing examples questions

What is AI washing?: AI washing is the practice of using AI-related terms—such as 'AI-powered', 'machine learning-driven', or 'intelligent'—in marketing without substantiating what the AI actually does, how it was tested, or what evidence supports the claimed result. Regulators including the FTC (Operation AI Comply, September 2024) and the SEC (Delphia and Global Predictions enforcement, March 2024) have taken action against companies whose AI claims were not backed by the evidence the wording implied.
What makes an AI claim 'high risk' for buyers?: An AI claim carries a higher evidence burden when it uses absolute or comparative language (zero, perfect, always, highest), professional-equivalence wording (like a lawyer, as good as a doctor), compliance certification language (HIPAA-compliant, GDPR-ready, fully ADA compliant), or quantified outcomes (saves 40% of costs, 99% accuracy) without naming the test conditions, populations, and failure modes that define the figure.
How is AI washing different from ordinary marketing exaggeration?: Ordinary puffery—words like 'leading', 'innovative', or 'cutting-edge'—is generally not actionable because it is understood as subjective opinion. AI washing claims involving measurable accuracy, safety outcomes, professional equivalence, or legal compliance are held to a higher standard: the FTC requires competent and reliable evidence to exist before the claim is published. Specific AI performance or compliance claims are not treated as puffery.
How do I spot AI washing when evaluating a vendor?: Look for five patterns in the marketing copy. Accuracy numbers without a named test set or error rate ('98% accurate'). Compliance labels without the auditor, scope, or what the certification does not cover ('HIPAA-compliant'). Professional-replacement phrasing without task limits or a human-review boundary ('like a lawyer'). ROI or cost-saving claims without a baseline, sample size, or industry scope ('saves 40% of costs'). And 'AI-powered' or 'AI-native' language without naming which workflow step the model runs, what input it takes, or what a person still reviews. Each pattern means the vendor should be able to provide specific evidence when asked—it does not mean the product cannot deliver.
What are common AI washing red flags in vendor marketing copy?: The most common patterns from FTC and ASA enforcement records: specific accuracy figures without a visible benchmark or false-positive rate; automated compliance wording that does not name what automation cannot cover; ROI guarantees without a comparison baseline or sample size; 'first,' 'only,' or 'best' claims without naming the comparison set or date; and AI product descriptions that list features rather than tested outcomes with stated limits. These are the wording patterns mapped in the signal catalog on the Methodology page.
Which companies have faced regulatory action over AI washing claims?: FTC enforcement cases with AI claim findings include Content at Scale AI (AI detector accuracy, 2025), accessiBe (automated accessibility compliance, 2025), DoNotPay (AI lawyer claims, 2025), Workado (AI content detector, 2024), Evolv Technology (AI security screening accuracy), and IntelliVision (bias-free facial recognition claims). The SEC took action against Delphia and Global Predictions in March 2024 over AI investment capability claims. ASA in the UK ruled against PixVideo AI over AI video advertising in 2024. These are official-source enforcement examples, not a company ranking or judgment.
What does an AI washing claim pattern look like in a real case?: One documented pattern: Content at Scale AI marketed an AI content detector with a specific accuracy figure (FTC case, 2025). The evidence gap was that the stated rate did not publicly name the content types, model versions, or mixed-authorship cases covered, and did not disclose a false-positive rate. Claim type: Accuracy / Performance. Wording signal: detector accuracy stated as a percentage without benchmark scope. Buyer question to ask: 'What exact content types and model versions were included in the test that produced this number, and what is the false-positive rate for each?'
How do I check an AI vendor's claims before relying on them?: For each AI claim on the vendor's public page: identify the exact words that raise the evidence burden (percent figures, compliance labels, automation language, or superlatives); categorize the claim type (accuracy, compliance, automation, ROI, or vague AI-powered); name the evidence that would support the claim in your specific use case; and write one concrete question to ask the vendor. Pasting the vendor's public page URL or the claim text into the checker on this site steps through this process and outputs a Claim Receipt you can use in vendor conversations or add to procurement notes.