Algorithmic persuasion is influence that learns who you are while it is influencing you.
Persuasion used to arrive as a message with an identifiable speaker. Algorithmic systems change the arrangement. They observe behaviour, infer likely responses, and continuously adjust what a person sees next. The result is not simply a persuasive argument delivered more efficiently. It is a changing environment designed around the probability of a response.
This is why recommendation feeds, targeted advertising, and automated ranking can influence people without presenting an explicit case. The system chooses what becomes visible, what follows it, and what is made easy to repeat. Personalisation turns a general message into a sequence of interventions fitted to one person’s habits, vulnerabilities, and predicted interests.
The important distinction is between assistance and persuasion. A recommender can reduce the cost of finding something useful, but the same infrastructure can optimise for attention, engagement, or conversion instead of the user’s stated interests. The system may then reinforce a preference, manufacture a new one, or narrow the range of choices that feels available.
Algorithmic persuasion is therefore a problem of agency as much as communication. It operates through algorithms, choice architecture, and manipulation at the same time. The question is not only whether a message is true or compelling. It is who controls the path by which the message reaches the mind.
Source: Institutional and academic synthesis: Malte Dold, “Algorithms and Autonomy: Regulating Recommender Systems in the Age of Hyper-Nudging,” Behavioural Public Policy Blog, 2025, drawing on Karen Yeung’s work on hyper-nudges.