The FTC vs. '98% Accurate' Claims: Lessons from the Workado Settlement

When the Coin Toss Beat the AI

In May 2025, the Federal Trade Commission reached a landmark settlement with Workado LLC that fundamentally changed how AI detection companies can market their products. The devastating finding: Workado’s “AI Content Detector,” marketed as “98% accurate” at identifying AI-generated text, actually performed at just 53% accuracy in independent testing-barely better than flipping a coin.

“Consumers trusted Workado’s AI Content Detector to help them decipher whether AI was behind a piece of writing, but the product did no better than a coin toss,” said Chris Mufarrige, Director of the FTC’s Bureau of Consumer Protection.

This was the first major enforcement action specifically targeting AI detection accuracy claims.

The Case: Marketing vs. Reality

Workado’s AI Content Detector launched in 2022, claiming superiority over “outdated” competitors by identifying content from ChatGPT, Claude, Gemini, and other models. Their marketing repeatedly claimed “98% accuracy” across multiple channels, targeting schools and businesses.

The FTC’s investigation revealed:

• Independent testing showed only 53% accuracy on general-purpose content
• The model was primarily trained on academic content, despite general marketing
• Significant false positive rates flagged legitimate human writing as AI-generated

Under the settlement, Workado must cease making unsubstantiated claims, retain documentation for future performance claims, notify affected consumers, and submit annual compliance reports to the FTC for four years.

Part of a Broader Crackdown

The Workado case emerged from Operation AI Comply, the FTC’s September 2024 enforcement sweep targeting deceptive AI claims. Other notable cases included:

• DoNotPay: The “world’s first robot lawyer” lacked adequate legal training
• Evolv Technologies: Falsely claimed AI-powered metal detectors could distinguish weapons from harmless objects
• IntelliVision Technologies: Claimed bias-free facial recognition trained on “millions of images” when actually trained on only 100,000 faces

Why 98% Was Impossible

AI detection faces fundamental challenges that make such high accuracy claims unrealistic. First, model evolution means AI continuously improves, making detection patterns obsolete. Second, domain specificity creates issues since academic, creative, and technical writing have different patterns. Human variability also poses problems, as non-native speakers may write in AI-resembling patterns. Finally, mixed authorship through human-AI collaboration creates ambiguous cases that are difficult to classify.

The FTC found Workado’s model was “only trained or fine-tuned to effectively classify academic content”-a critical limitation never disclosed to customers.

The New Compliance Reality

The settlement established new standards requiring companies to provide scientific substantiation with peer-reviewed testing methodologies. They must use representative datasets that span diverse content types and obtain independent third-party validation of their claims. Additionally, companies must provide clear disclosure of any training data limitations and performance variability in their systems.

Organizations evaluating AI detection tools should take several key precautions. They need to demand independent validation from recognized testing organizations before making purchasing decisions. It’s critical to require pilot testing on representative content that matches their actual use cases. Organizations should also insist on performance guarantees with clear remediation terms if accuracy falls below promised levels. Finally, they must establish robust monitoring systems to track actual deployment accuracy over time.

The Blurring Line: Why Detection Misses the Point

The Workado settlement exposes a deeper truth: by 2025, the clean distinction between “AI-generated” and “human-generated” content has largely disappeared. The real world isn’t binary-it’s a spectrum of human-AI collaboration.

Consider how professionals actually work with AI today:

• Writers use AI for brainstorming, outlining, and editing assistance
• Developers leverage AI for code suggestions, debugging, and documentation
• Students employ AI tutors for learning and concept clarification
• Researchers use AI to analyze data patterns and generate hypotheses

This isn’t “cheating”, it’s augmented human intelligence. Trying to draw precise lines between “human” and “AI” work is like distinguishing between thoughts influenced by books you’ve read versus original insights.

What Really Matters: Ethics Over Origins

The Workado case teaches us that obsessing over detection accuracy misses the fundamental question: Are we creating value ethically and transparently?

Instead of worrying whether AI-assisted work will be “detected,” content creators should focus on fundamental questions of integrity. They need to consider whether they’re being transparent about their process when transparency is called for. They should examine if they’re truly adding genuine value through their expertise and judgment. Most importantly, they must ensure their final product is accurate, helpful, and honest to their audience.

Here, for example, I clearly use AI models as part of my post drafting and editing process, and have published some of my tools (like writing-tools-mcp) as open source projects.

Rather than getting caught up in an endless cycle of detection and evasion, educators and organizations need to take a more constructive approach. This means prioritizing AI literacy and teaching ethical usage principles. It involves redesigning assessments to emphasize critical thinking skills over mere content production. Organizations should develop and communicate clear policies about acceptable AI collaboration. The focus should shift toward measuring and improving quality outcomes rather than policing the methods used to create content.

Conclusion: Building an Ethical AI-Augmented Future

The Workado settlement signals a maturation of how we think about AI in society. The era of binary “human vs. AI” thinking is ending, replaced by a nuanced understanding of human-AI collaboration.

The Real Lessons for Practitioners

1. Ethical Marketing Matters: Honesty about capabilities and limitations builds long-term trust
2. Focus on Value Creation: Energy spent detecting AI usage should be invested in ensuring high-quality, ethical output
3. Transparency Is Key: Appropriate disclosure protects both creators and audiences
4. Embrace Augmentation Thoughtfully: Use AI wisely within ethical and regulatory bounds

The future belongs to those who can work ethically and effectively in an AI-augmented world. This means developing AI literacy, building systems that prioritize quality and ethics over detection, and focusing on outcomes that benefit society.

In 2025 and beyond, the question isn’t whether AI was involved in creating something-it’s whether that something is actually good.

Sources and References

• Crowell & Moring: “FTC Cracks Down on AI Model’s AI Detection Claims”
• National Law Review: “Workado Settles With FTC Over Alleged Misrepresentations of AI Accuracy”
• PCMag: “Did You Use This AI-Detection Tool? The Results May Be Bogus”
• Lathrop GPM: “Transparency and AI: FTC Launches Enforcement Actions Against Businesses Promoting Deceptive AI Product Claims”