AI slashes persuasion costs-and shifts power to those with data and distribution

Generative models have made it orders-of-magnitude cheaper to produce plausible, tailored messages-text, image, and voice-on demand. Under the hood, the stack is familiar: LLMs for copy, synthetic media for assets, and automated orchestration to A/B test variants with bandit-style optimization. What’s notable here isn’t a new desire to persuade, but a changed cost curve: marginal content is near-free, iteration cycles collapse from days to minutes, and “creative” becomes a function of compute and data. The bottleneck moves from crafting messages to acquiring high-quality targets and reliable distribution.

The bigger picture: benefits accrue to actors with first-party data and reach-platforms, major publishers, large brands, and well-funded campaigns. Indie operators get better tools, but the leverage sits where identity graphs, lookalike modeling, and delivery pipes already exist. Worth noting, detection remains hard; provenance (C2PA, content credentials) and rate limits are more promising than after-the-fact classifiers. Expect ad marketplaces to price in authenticity signals and platforms to tighten API access, throttle automation, and require disclosures. This isn’t mind control-effects are typically marginal and context-dependent-but marginal effects at scale compound. In competitive markets (and close contests), shaving a few basis points via faster, cheaper experimentation is material. What’s actually new is the unit economics and speed; the hype is assuming universal persuasion. The strategic constraint is shifting from message creation to governance of data, distribution, and trust.