Author: Nita Farahany Source: https://nitafarahany.substack.com/p/the-persuasion-exchange Published: March 2, 2026
"The Persuasion Exchange" is an essay by Nita Farahany, published March 2, 2026, that opens Week 7 (Persuasion / Personalized Manipulation) of her Advanced Topics in AI Law and Policy course (Class 7 of roughly 30). It anchors on Coffee v. Google (the Pixel-buds personalized-recommendation case) and the FTC's loot-box enforcement record, and frames the question of when AI-driven personalized persuasion crosses from advertising into manipulation, and which legal regime, if any, reaches the conduct.
Summary of argument
Farahany asks at what point personalized persuasion crosses from advertising into manipulation, and which legal regime reaches it. To structure the question she sets out a personalization gradient, tests where on that gradient persuasion becomes manipulation using a two-conditions test introduced earlier in her intro course, and surveys the consumer-protection, securities, and election-law regimes that touch personalized persuasion, concluding that none reaches the cross-domain case of closed-loop personalized persuasion.
Key claims
The personalization gradient runs across five points on a spectrum:
- Mass advertising — the same message to everyone.
- Demographic targeting — different messages to different cohorts.
- Behavioral targeting — messages tailored to inferred interests from browsing and purchase history.
- Real-time personalization — messages adapted to current emotional state, location, and recent activity.
- Closed-loop optimization — the system learns from each individual's responses and adapts the persuasion vector continuously.
On where persuasion crosses into manipulation, Farahany returns to the Class 15b (intro course) two-conditions test: the system pursues an objective, and the system alters decision-making covertly. She argues that most personalized advertising satisfies condition 1 but not condition 2, because the persuasion is not exactly covert: people know ads exist. Closed-loop optimization satisfies both conditions, learning a user's specific vulnerabilities and exploiting them with content the user cannot easily recognize as adapted to them.
The essay uses two bodies of legal material to ground the gradient. Coffee v. Google (D. Mass., 2023) involved a plaintiff alleging that Google's Pixel-buds product page used personalized recommendations and pricing optimized for the individual user, in violation of Massachusetts consumer-protection law. The case survived a motion to dismiss and turned on whether personalized pricing-and-recommendation is "deceptive" when the personalization itself is not disclosed; it settled in early 2026. The loot-box enforcement record comprises FTC v. Epic Games (2022, $245M), addressing Fortnite's V-Buck system, dark-pattern checkout, and design choices targeting children, which settled; FTC v. Activision Blizzard (ongoing), addressing microtransaction practices in Diablo IV; and the treatment of loot boxes as illegal gambling in Belgium and the Netherlands.
Farahany identifies a legal-regime gap across three regimes that each touch personalized persuasion. Consumer protection (FTC §5, state UDAP) reaches deception but struggles with personalization-as-architecture. Securities law reaches financial-product persuasion only and is not generalizable. Election law reaches political persuasion only, and the FEC has refused to extend it to AI-personalized political advertising. By her account, none of the three reaches the cross-domain phenomenon of closed-loop personalized persuasion.
Relationships
- part-of: Nita Farahany Advanced Topics course (Class 7 of ~30)
- related: Two Conditions and Three Weapons of AI Manipulation, Five Paradigms of AI Manipulation Governance
- previous: Inside My Advanced Topics Class 6.3: Your Brain Everywhere (Farahany, February 2026) next: Inside My Advanced Topics Class 7.2: The Statute That Couldn't Stretch (Farahany, March 2026)