Auditing AI-Washing in German Startups: A Mixed-Methods Study of Marketing Claims and Perceptions

URL SCAN: Auditing AI-Washing in German Startups: A Mixed-Methods Study of Marketing Claims and Perceptions
FIRST LINE: # Computer Science > Computers and Society

The Dissection

The paper audits a symptom of AI’s promotional economy: founders use “AI” as credibility collateral, investors reward branding they cannot verify, and consumers accept claims in opaque domains. Its 100-startup annotation study and 63 expert interviews turn exaggeration into an auditable compliance problem.

That is useful forensic work—but it examines the wrapper around AI, not AI’s power to reorganize production.

The Core Fallacy

The paper treats inaccurate claims and disclosure gaps as the primary defect. Under the Discontinuity Thesis, even perfectly truthful, fully disclosed AI can sever the mass employment–wage–consumption circuit. A startup can pass every audit and still automate cognitive labor, concentrate ownership, and make human participation economically redundant.

Disclosure may improve informational quality. It does not defeat Cognitive Automation Dominance, Coordination Impossibility, or Productive Participation Collapse. The proposed remedy cleans the market’s claims while leaving the underlying displacement mechanism intact.

Hidden Assumptions

• Investors can be disciplined into reliably verifying AI claims rather than rewarding whatever produces superior returns.
• Consumer distrust will correct the market instead of merely filtering out weak sellers while strengthening credible automation.
• Disclosure requirements can keep pace with rapidly changing systems and opaque performance claims.
• Matching the claimed capability to the underlying system makes deployment sufficiently legitimate.
• Ethical concern and public trust are central constraints on the transition, rather than friction that capital can eventually route around.
• A better-informed market remains compatible with mass human economic participation.

Social Function

Classification: partial truth, transition management, and ideological anesthetic.

The paper correctly identifies promotional deception, unverifiable performance numbers, unacknowledged limits, and ethically concerning uses. But by framing the problem as AI-washing, it makes the crisis appear governable through audits, disclosure, and consumer protection. That preserves the legitimacy of the market by relocating the danger from automation itself to dishonest marketing.

Its reforms would, if successful, produce cleaner and more credible AI deployment. That reduces reputational friction around the very systems capable of accelerating labor displacement.

The Verdict

A competent audit of AI theater, strategically blind to AI’s structural consequence. It can distinguish marketing fraud from demonstrated capability, but it cannot preserve the post-WWII economic order once capable systems replace economically necessary human labor. The paper cleans the mask; it does not stop the machine. AI-washing is a transient sales pathology around the terminal threat, not the terminal threat itself.