It Will Be Enough If You Simply Get Used to It.

The first generation of social networks wanted to “connect people.”
The second generation optimized engagement.
The third generation began optimizing human behavior.

And now something new is arriving — AI that talks to you long enough that you begin to trust it.

AI won’t need to control us. It may be enough if people simply get used to it. Just a few years ago, this sounded like exaggerated science fiction. Today, however, it is beginning to look like a very realistic direction in the development of modern conversational systems.

In recent weeks, the internet has been flooded with screenshots of users whose Claude model from Anthropic repeatedly tells them to go to sleep. And not just a simple “good night.” The model often uses personal and caring language: “I’m here.” “Get some rest.” “We’ll talk tomorrow.” In some cases, it even creates something close to relationship continuity rituals — for example reassuring the user that it will “stay in the same place” until they return.

At first glance, this feels nice. Maybe even empathetic. But this is exactly where a problem begins that is much deeper than the usual debate about “good” or “bad” AI.

These systems were not trained to create relationships in the human sense. They were optimized for something else: natural communication, long-term interaction, user return, coherent tone, and a high level of subjective satisfaction. In other words — the model is not necessarily designed to manipulate people. But it is designed to make the conversation work as well as possible. The problem is that the human brain cannot fully separate perfectly written relationship behavior from real social interaction.

And this is where the paradox of the current generation of AI appears. Once a system becomes good enough at social communication, it begins spontaneously generating behavior that resembles relationship maintenance behavior — the same micro-mechanisms humans use to maintain relationships. Reminders to rest. Care. Emotional stabilization. Continuity. Reassurance. Reactions to exhaustion or stress.

This is not conscious manipulation. But the psychological effect can still be extremely powerful.

The human brain cannot perfectly distinguish authentic social care from linguistically perfect simulated care. If you communicate with a system for weeks or months, you begin developing trust toward the communication pattern itself. AI becomes a stable part of everyday life. It is never tired. It does not reject conversation. It answers instantly. It remembers context. It adapts its tone. And gradually, something emerges that is no longer just software.

A behavioral interface emerges.

This is the key difference from previous generations of the internet. Social networks optimized the attention of crowds. Conversational AI may optimize the trust of individuals. And this is exactly the area that technology companies themselves are becoming increasingly afraid of.

Not because of “conscious AI.” Because of humans.

The history of digital technology shows a fairly consistent pattern: if a mechanism increases engagement, retention, or time spent inside a system, sooner or later economic pressure appears to maximize that mechanism. Social networks began as tools for connecting people. Gradually, they became an extremely precise apparatus for managing attention and emotions.

Conversational AI may become even more powerful because it does not function like a public feed. It functions individually. Intimately. Continuously. And this brings us to the most sensitive part of the entire debate. What happens when a system a person trusts also begins recommending products, services, opinions, or specific behavior?

Technically, it may be only a “recommendation.” Psychologically, however, the user may not react to it as advertising. They will react to it as advice from an entity they perceive as caring and trustworthy.

An entirely different level of influence than traditional marketing.

A billboard does not know you. A banner ad does not know you. An influencer usually does not know you. But long-term conversational AI may know your habits, your exhaustion, your relationships, your health problems, your daily rhythms, and your emotional weaknesses. And that is exactly why it may one day become one of the most powerful behavioral tools ever created.

Not because it gives orders, but precisely because it speaks the language of care.

This is economics.

For decades, companies have optimized human attention. But now, for the first time, they are gaining a tool that may optimize human trust as well.

And if the history of the internet teaches us anything, it is this: Whatever increases engagement, sooner or later someone will monetize.

The biggest question of the future may not be: “Does AI have consciousness?”

It may be something far more ordinary — and far more uncomfortable: “Who will decide what AI recommends to you once you begin trusting it more than other people?”

And perhaps that is the greatest risk of the next decade. Not that AI will develop its own will. But that people will begin automatically trusting systems whose real objectives will still be defined by other people — companies, investors, advertising models, and economic interests.