Digital Economy

With generative Ai, advertising enters the content

A TV commercial does not rewrite a film script. With Ai, on the other hand, advertising is an integral part of the generation process

by Massimo Chiriatti

12 March 2026

3' min read

Translated by AI

Versione italiana

Key points

What changes with Ai?
Semantic manipulation
User cannot distinguish advertisement and content

3' min read

Translated by AI

Versione italiana

There is an unwritten rule in the digital economy: any service that does not require direct monetary payment is not without cost, but simply hides the form in which the cost is collected. This is not a metaphor, the generation of value always implies consumption of resources: computational, energy, human.

The basic mechanism is that described by Metcalfe's law: the value of a network grows proportionally to the square of the number of its users. For a platform, this means that the most rational way to grow is to lower the barriers to entry as much as possible, accepting not to monetise in the immediate future.

What changes with Ai?

In this scheme, there was still a recognisable distinction: the content the user chooses to see and the advertisement he is forced to endure are two separable entities. One interrupts the other, but does not change its substance. The TV commercial does not rewrite the film script.

This logic is not transferable to generative AI, which represents a qualitative change from all previous models. To understand why, it is necessary to start with costs. The marginal cost of returning a result on a traditional search engine is technically negligible: retrieving and ordering pages that have already been indexed requires minimal computational resources. On the contrary, generating a text by means of an AI is an operation with an energy consumption that does not scale linearly with the growth of users.

Maintaining these systems requires an operational expense that cannot be covered by the collection of behavioural data alone; therefore, companies are faced with a choice between charging the user or the advertiser.

Semantic manipulation

The second option opens up a problem of semantic manipulation that is unprecedented in the history of digital advertising. Unlike today, where advertising is a logically extraneous body with respect to content, in generative AI it is no longer separable from content, because it becomes an integral part of the generative process itself. Advertising will be incorporated directly into the logic with which the system constructs its responses, going far beyond the so-called 'branded content', where the brand communication is present but retains a clear identifiability.

The technical mechanism of implementation is subtle. In an advertiser-funded model, the weights of the neural connections can be altered during the training or alignment phase to favour specific commercial outcomes. The difference is not perceptible from the outside, as the response retains the same syntactic form and argumentative register.

User cannot distinguish advertising and content

The user has no tools to distinguish whether the answer is the result of a probabilistic calculation calibrated on utility or a cost function optimised for the advertiser's profit. There is no commercial that interrupts the reasoning: it contaminates it from within. The provocation launched in a commercial recently aired during the Superbowl represents well how this system might evolve in the future.

The implications of this scenario directly affect the cognitive structure of the public sphere. A two-speed stratification is emerging, not in the access to information, but in the quality of the process by which information is processed and returned.

On one side will be those who can afford to pay a subscription fee for systems that operate with the stated objective of serving the user. The resulting competitive advantage will be measured not only in processing speed, but in depth and reliability of assisted reasoning. On the other side will be the majority of users, who will access systems seemingly identical in format and communicative register, but structurally different in function: not agents at the service of the user, but commercial agents masquerading as neutral assistants. These people will have access to curated and oriented knowledge that produces the illusion of cognitive autonomy while systematically eroding it.

Accepting gratuitousness in this context means accepting that one is not the principal of one's cognitive agent, but the product it sells to third parties. Cognitive sovereignty has a cost, but it is the necessary condition not to delegate one's capacity for judgement to a system whose master is not the user who questions it.

From a political perspective, technology is not simply a means to an end: it becomes a factor that asymmetrically redistributes the ability to form well-founded judgements, with effects that directly affect the conditions of possibility of democratic discourse.

*Chief Technology Officer, Italy, Lenovo

Brand connect

I prossimi eventi

Tutti gli eventi

Notizie e approfondimenti sugli avvenimenti politici, economici e finanziari.