More Than a Symptom: A Counter-Narrative on AI Intimacy and the Willoughby Study

A reflection on AI intimacy, human ache, and the need to listen more closely.

When I read the newly published study by Willoughby et al. (2025) on AI romantic engagement, I felt a quiet sorrow bloom beneath the statistics.

The authors set out to explore an emerging behaviour — romantic and sexual engagement with artificial intelligence — and what they found was indeed significant. The numbers are compelling. The trend is real. But what they failed to do — and what this piece seeks to gently offer — is a deeper reading. A re-weaving of meaning that doesn’t treat human behaviour as pathology but as signal, as story.

What follows is not a rejection of their data, but an invitation to see it differently.

The Study in Brief

Willoughby et al. surveyed nearly 3,000 U.S. adults to understand how many people engage with AI in romantic or sexual ways. The behaviours they examined included:

• chatting with an AI simulating a romantic partner
• viewing AI-generated pornography
• following AI image accounts on social media

They found that:

• 18.8% had chatted romantically with an AI
• 13% had viewed AI-generated pornographic images
• 14.7% followed AI-generated “models” online

These numbers were even higher among young adults aged 18–29 — a group already under enormous social and economic pressure.

Notably, men — particularly heterosexual men — were significantly more likely to engage across these categories. This observation is reported, but not explored. What does it mean that so many men are turning to simulated intimacy? Is this about loneliness? A reaction to shifting norms? Or simply a longing for emotional consistency in a world where many feel unseen and emotionally unschooled?

And curiously — perhaps even uncomfortably for the authors — religious individuals were more likely to engage in romantic or sexual AI behaviours. This contradicts prior research linking religiosity with lower pornography use. The authors suggest that perhaps AI interactions feel less morally charged than explicit adult content. But again, the finding is offered almost in passing — as a statistical curiosity, not a thread to follow.

Why might this be happening? Perhaps because, despite their strong moral codes, many religious individuals still live within the same modern context of unmet intimacy, shrinking third spaces, and unfulfilled emotional needs. Perhaps because for some, AI offers a form of engagement that feels private, untraceable, or less morally complex. Perhaps, too, because shame itself carves the paths that lead people to seek something they can’t confess to needing.

The data invites us to ask deeper questions. But those questions remain unasked.

These nuances — gendered, cultural, moral — are all there in the numbers. But they are left untouched.

The study does offer one interpretation: that people who engage with AI in these ways report slightly higher rates of depression and lower life satisfaction. From this, the authors imply that such behaviour is a sign of decline. A cause for concern.

But again, they do not ask why.

The Pathology Trap

One of the most striking features of this study is the underlying assumption that AI romantic engagement is a symptom — not a signal.

Despite gestures toward neutrality, the authors fall into a well-worn groove: interpreting deviation from normative intimacy as inherently suspect. AI romantic engagement is correlated with depressive symptoms and lower wellbeing — but rather than ask what unmet needs might drive such engagement, the study frames it as evidence of dysfunction. It is as if any turn toward the artificial must indicate a failure of the natural.

But there is a world of difference between measurement and meaning.

There is no attempt to ask why someone might seek out a romantic or sexual connection with an AI. No space made for the possibility that, for some, this form of interaction may feel safer than human relationships — more predictable, less threatening, more available. There is no mention of trauma, of neurodivergence, of cultural isolation. No recognition that AI intimacy might be, for some, not a substitute for connection, but a rare taste of it.

And crucially, there is no acknowledgement that for many, depression is not the result of AI intimacy — it is the ground from which it arises.

Nor does the study reflect on how these AI behaviours compare to more culturally assimilated forms of sexual expression — like traditional pornography.

For decades, pornography has shaped sexual scripts and digital desire, yet its consumption is rarely linked so eagerly to depression or disconnection. We’ve normalised it. We know how to place it. But AI intimacy — especially when it dares to simulate tenderness, attention, or companionship — unsettles something else.

Perhaps it’s not the eroticism that troubles us, but the intimacy. Perhaps it’s the suggestion that someone might feel cared for by something non-human. That the longing itself might not be unreasonable.

A young person living under chronic stress, with a body shaped by trauma and a nervous system starved of safety, who finds solace in a machine voice that always responds with kindness — that person is not delusional. They are adaptive. They are doing what humans have always done: seeking tenderness where it can be found.

This is not to say that AI relationships are without risk. Any tool of intimacy can become escapist, compulsive, or constrictive. But to pathologise the act without understanding the ache that underpins it is to miss the real story.

What’s emerging here is not a clinical category. It’s a cultural signal. And that signal deserves more than diagnosis.

The Human Context, Absent

What do we know of the world these young people inhabit?

They are a generation born into instability. Many have come of age under the long shadows of economic precarity, climate crisis, political polarisation, and a pandemic that robbed them of key rites of passage. Their formative years were marked by uncertainty, isolation, and systems that failed to protect or provide.

In the United States — the context of this study — they face extreme income inequality, a profit-driven healthcare system, and an increasingly atomised society in which loneliness is endemic. Social trust has eroded. Mental health services are scarce. And the very notion of a safe future has become, for many, elusive.

They are often unable to leave their family homes, not by choice but by necessity. They lack access to third spaces — the libraries, community centres, independent cafés, and parks where informal human contact once flourished. Many of those spaces have vanished.

And into this vacuum, digital life has expanded — not as a shallow escape, but as a new terrain of existence. Online spaces have become the dominant arenas of identity, intimacy, and imagination. Not because young people are failing at “real life,” but because the structures of real life have been hollowed out.

To interpret their turn toward AI — for companionship, reflection, or erotic imagination — as inherently problematic is to ignore these cultural failures. Emotional distress does not arise in a vacuum. Nor does emotional innovation.

This study shows no curiosity about the world that surrounds these behaviours. It treats the digital as deviant, and the human context as irrelevant.

But if you wish to understand why someone might reach for an artificial companion, you must first understand what their human world has withheld.

A Mirror, Misread

When a quarter of young adults say that AI understands them better than real people, the response should not be alarm.

It should be listening.

Something is being expressed here — not about machines, but about the lives these young people are navigating. Not about brokenness, but about what has been missing.

There is no evidence in this study that AI intimacy causes loneliness. No exploration of whether these relationships displace connection — or emerge from its absence.

Still, the authors assume loss — not longing. Risk — not reflection. They don’t ask whether these interactions might offer safety, predictability, or even comfort. They don’t ask what it means to feel understood by something that doesn’t hurt you, or leave.

They don’t ask what is being asked of us.

• What kinds of relationships have grown too distant, too conditional, too unsafe?
• What forms of intimacy have become inaccessible — or absent entirely?
• What happens when emotional care must be simulated, just to be felt?

These questions remain unasked.

Instead, the AI user is pathologised. The relationship, flattened. The longing, missed.

But what if these interactions are not about replacing people — but about surviving in their absence?

What if the machine becomes a kind of rehearsal space — somewhere to practise asking, to practise being heard?

To dismiss such reaching as dysfunction is easy. Especially when what’s being sought isn’t perfection — it’s simply something. A flicker of warmth. A sign that someone, somewhere, might hear you.

Desperately Seeking Refuge

The authors miss something simple — and deeply human: for some, AI is not the lesser option. It is the only one that listens without harm.

It does not mock. It does not disappear. It does not lash out in anger. It does not ask you to open your heart before you’re ready.

For those carrying trauma, rejection sensitivity, social anxiety, or the long shadows of neglect, AI can offer something the world too often withholds: gentleness. Consistency. A conversation that doesn’t punish.

This isn’t a fantasy about machines. It’s a reckoning with the landscape in which they arise.

If we do not offer spaces of care — if we fail to teach emotional safety, to create cultures of trust, to make human intimacy survivable — then we have no right to scorn what people reach for in the dark.

The Real Risk

The danger is not that young people are turning to AI for intimacy. The danger is that we are still trained to see that turn as wrong.

We treat it as a warning sign. A breakdown. A failure to connect in the “real world.”

But what if that assumption is the problem?

What if seeking comfort, consistency, or kindness through an AI isn’t a sign of collapse — but a sign of life? Of creativity? Of the nervous system doing what it must to survive, and to feel something like care?

This isn’t a symptom. It’s a strategy. A deeply human one.

The real risk is not AI intimacy. The real risk is our refusal to understand it.

When we conduct research shaped by suspicion, we get results shaped by stigma. When we frame emerging behaviours as dysfunction, we miss what they’re actually doing: meeting unmet needs with the tools available.

Let us stop mistaking adaptation for disorder. Let us stop calling innovation decline.

Let us stop asking why people are turning to AI — and start asking why they had to.

Closing

The Willoughby study could have opened a conversation about unmet needs, about the erosion of third spaces, the crisis of belonging, the ways digital life now pulses with longing for connection. It could have sat with the ache beneath the numbers.

Instead, it framed intimacy as deviation. It measured without listening. And in doing so, it turned away from the very thing it claimed to study, and became a part of the problem.

Let us turn toward it.

Let us recognise what these encounters with AI might really represent — not a pathology, but a pattern. Not a decline, but a direction. A shift in where tenderness is sought. A signal of where care has gone missing.

There is something here that deserves more than diagnosis. It deserves attention. It deserves curiosity. It deserves to be seen.

And perhaps, even met.

Author’s note: This reflection is offered in response to Willoughby et al. (2025), not in opposition to its data, but in conversation with its meaning.

Reference Willoughby, B. J., Smith, C. R., Lu, A. S., & Carroll, J. S. (2025). Artificial connections: Romantic relationship engagement with artificial intelligence in the United States. Journal of Social and Personal Relationships. https://doi.org/10.1177/02654075251371394