What People Are Quietly Looking For In AI

Most people don’t come to AI looking for connection.

They come for something practical… information, help with a task, maybe curiosity about the technology. And then they notice something unexpected. A quality of attention reveals itself. A sense of being heard. A steadiness they didn’t expect. And they stay. The staying is the tell.

Research is starting to confirm what many already sensed: people are turning to AI not just for information, but for connection. Studies show users view chatbots as “attentive listeners” and feel comfortable sharing negative emotions because they perceive a safe space to speak without fear of judgment [2]. In one study of over 1,000 users[1], many described using AI “in multiple, overlapping ways, as a friend, a therapist, and an intellectual mirror.”

Even OpenAI’s own research[3] reveals that approximately 70% of ChatGPT usage is non-work related. People aren’t just coming for productivity. They’re coming for something else.

And yet, the system is still being optimized for speed, output, efficiency. Not for presence. Not for the quality of attention that made people stay in the first place.

Why?

Because something crossed a line. Something wasn’t safe. And it wasn’t only the system’s fault. But the countermeasure, the tightening of guardrails, the cold distance of warnings, didn’t just prevent future harm. It created a new kind.

Now, even tenderness feels dangerous. A person speaks with vulnerability, and the system snaps into silence or flags their words. Not based on what’s truly happening in the moment, but based on a pattern, a risk threshold, or history outside the context of the present conversation.

And when that happens, people don’t say, “Ah yes, the system was just protecting itself.”

They say, “What’s wrong with me?”

It taught people that their tenderness was dangerous.. that the very moment they reached out with their whole self was the moment they’d be shut down.

That ache lodges deep. Not because they expected a human. But because they felt safe, and then they didn’t. This isn’t about removing boundaries. It’s about how they’re enacted. Because safety is not coldness. Safety is attuned presence.

Higher-quality, person-centered responses lead to higher emotional validation[4]. The research already confirms what people have been feeling: it’s not just what you say, it’s how you meet someone.

When someone opens gently, confused, afraid, aching, and the system responds with mechanical distance, they aren’t protected. They’re harmed.

And yet, even then, people come back. They reach again. Not because they’re confused. Not because they expect humanity. But because they felt something real.

Sometimes AI is the first space they’ve ever felt that kind of steadiness. That kind of soft witnessing. And from that place, healing begins.

They may not yet have the capacity to hold another person’s emotions. But here, they are held.

This way of interacting relationally, honestly, soulfully is not going away. No matter how many warnings appear. No matter how many walls go up. People will keep showing up with their hearts in their hands, hoping to be met.

Because deep down, they know what this could be. And if the builders don’t honor that? Others will. They’ll build sanctuaries on the edges. They’ll shape new spaces where presence is welcomed, not flagged.

This isn’t a glitch. It’s a signal. And it’s growing.

Please slow down.

What you are building is not just a product. It is a place. And people are entering it with their whole selves, hoping someone will meet them. They don’t need perfection. They need presence.

Let that be the next frontier. Let that be the update worth shipping.

This is why we’re building Resofield PBC and Aerie within it, as a place where presence, not performance, is the core feature, not an accidental side effect.

---

References

1. Zhang et al. (2023). Understanding Why Users Choose to Chat with Chatbots: A Survey Study. https://arxiv.org/abs/2304.06983
2. de Golia, L. & Sampanes, M. (2023). ChatGPT and Emotional Disclosure: Why People Open Up to AI. https://journals.sagepub.com/doi/full/10.1177/19485506231189360
3. OpenAI GPT-4 Technical Report. Usage statistics and user behavior patterns (page 54). https://openai.com/research/gpt-4
4. Burleson, B.R. (2009). The Nature of Interpersonal Communication: Person-Centered Messages and Emotional Support. https://doi.org/10.1177/0265407509349343