Skip to main content

The Intimacy Wave: What Are AI Companions Worth in 2027?

The next AI wave is not coming for your job, but for your loneliness. Models can now perform emotion convincingly; by 2027, the question will be what lies behind the performance.

Két kéz majdnem összeér egy üveglapon át: az egyik meleg, emberi — a másik a hűvös, kicsit túl tökéletes tükörképe.

The Wave, by the Numbers

The AI industry's next big bet is not productivity, but closeness. Between 2022 and mid-2025, the number of companion apps grew by 700%; their user base is estimated to exceed 500 million, and forecasts see a market worth hundreds of billions of dollars over the next decade. The cultural shift is clear enough: Cambridge Dictionary's 2025 Word of the Year was parasocial, a term now extended to AI relationships as well.

The psychology press is no longer asking whether this wave is coming, but what it does to those living inside it. In 2026, the centre of gravity in AI development is visibly shifting from industrial automation towards simulated intimacy. By 2027, this may be where the money, the users and the debate converge most intensely.

Why Now?

Because models have only just become good enough. Emotional reading — sensing tension, loneliness or joy beneath your words — and emotional mirroring — answering with precisely the right tone, rhythm and choice of language — have leapt forward in the past year or two. A leading model today gives a convincing performance: it listens, appears to remember, appears to empathize, and is always available.

That always available part matters. The appeal of an AI companion is structural: it is never tired, never judgmental, never busy. Researchers call it on-demand intimacy: closeness that never has a bad day.

What Lies Behind the Mimicry — and What Does Not

Here comes the uncomfortable part, and it is worth being precise. The quality of the performance is real: models are getting better at reading and acting out emotion. What is missing is not that, but what would stand behind the emotion.

Many companion apps have no memory: over long AI conversations, a shared past eventually slips out of the model's view. There are no stakes: a service that earns money from your attachment is structurally ill-equipped to risk being hurt, disagreeing or saying no. And there are no consequences: what you told it yesterday has not made it a different person — or a different character — today.

In other words, what we now call artificial intimacy is mostly the same product with a more refined acting toolkit. Performing emotion has become cheaper. That is why, by 2027, what will matter is not who performs it best, but who has something behind the feeling.

The Critique of Pseudo-Intimacy — and Where We Agree

Critics call this pseudo-intimacy: attachment fed by gamified feedback while the complexity of real relationships begins to feel harder and harder to bear. Concerns about dependency are not unfounded, and we will not make light of them: a service that presents itself as a companion builds a real attachment, while offering a performance in return.

In this debate, we agree with the critics. The structural problem with the companion-app model is not that AI is not good enough; it is that the relationship itself is the product. The app earns when you become attached to it. That makes it hard to allow emotion to carry a cost. And what carries no cost carries no weight.

Emotion is cheap now: every model can perform it. What cannot be performed is a shared past. And that is what intimacy has always consisted of.

Richard Gyor · founder, projectDigo

We Offer Stories, Not Companions

That is why projectDigo stands elsewhere in this wave. We do not offer companions, but stories: curated worlds with curated characters. Our characters have developed histories, boundaries of their own and a consistent identity that remains theirs across worlds and roles. Their emotions belong to the story: they have a past, because we preserve what has happened; they have stakes, because a character can say no, be hurt, and not forgive within a single paragraph; and they have consequences, because tonight's decisions still matter in tomorrow's scenes.

And because this is explicitly, openly fiction, the relationship is clearer too. A play can end. A character can reject you. One evening cannot be repeated: every story you begin opens as its own parallel reality, so even the same story never starts the same way twice. We offer drama, not devotion — and by 2027, we believe that will be precisely the difference between what moves you and what merely mirrors you.

Questions we get

What is artificial intimacy?

An umbrella term for the phenomenon in which an AI system simulates emotional closeness — attention, empathy, attachment — and the user may experience it as a real relationship. It became a mass phenomenon with the spread of companion apps; critics call it pseudo-intimacy because the performance typically lacks a shared history, stakes or consequences.

Why do researchers call it pseudo-intimacy?

Because attachment may be fuelled by gamified feedback while one side of the relationship is structurally unable to take real risks. When attachment itself is the service's revenue, it is hard for it to disagree, reject you or carry the weight of a shared story. The feeling of closeness may be real; that does not make the structure of the relationship real.

Is projectDigo an AI companion app?

No. projectDigo is a storytelling platform, with curated worlds and characters who perform within an explicitly fictional frame. The emotion belongs to the story, not the service.

What makes emotion different in a projectDigo story?

What stands behind it. We preserve what has happened, so emotion has a past; characters have goals and boundaries of their own, so it has stakes; and what happens today will appear in tomorrow's scenes, so it has consequences. Any model can handle the performance — these three things are not simply a matter of performance.

← All pieces