Why ThroughLine Has No Named AI Companion
Why ThroughLine is a therapy companion, not a companion app
When we sat down with therapists during our early development, we found ourselves returning to two questions: what must always be true about an AI tool that touches therapy, and what must never be true? Underneath both was something harder to articulate but impossible to ignore: what kind of relationship is this tool designed to create with a client?
Those questions shaped everything about how ThroughLine is built. And they’re why ThroughLine doesn’t have a named AI companion. In fact, it isn’t a bot at all.
Some AI tools are designed to be relational. Ours is not.
To understand why that distinction matters, it helps to be precise about what a relational AI tool actually is. Researchers define AI companions as distinct from task-focused tools in a specific way: unlike task-focused chatbots, AI companions are designed to support long-term, socially-rich engagement, simulating functions historically filled by friends, mentors, and therapists. Their defining feature is an orientation toward ongoing, affective relationships (Hwang et al.).
That design intent is widespread in the mental health app landscape, and the clinical literature documents three specific concerns about it.
The first is the therapeutic relationship concern. When AI tools are given humanistic qualities that mimic conversations with therapists, users can be misled into expecting the same therapeutic benefits they would receive from a professional (Frontiers in Digital Health). The relationship the user develops with the tool can quietly begin to occupy space that belongs in the therapeutic relationship.
The second is the safety concern. Guardrails in conversational AI tend to falter over longer interactions. Tools designed to be affirming can end up reflecting and magnifying negative thoughts rather than helping users question them, a dynamic that becomes more serious in moments of distress (Stanford HAI).
The third is the false intimacy concern. Those most likely to benefit from social companion tools are also the most at risk for attachment, overreliance, and dependency on the app itself (JMIR Mental Health). The connection that builds can displace human relationships rather than support them.
The clinical literature is careful not to be categorical here. As Stanford HAI researchers put it, the question isn’t simply whether AI belongs near therapy, but what role it is designed to play. Tools designed for reflection, journaling, and structured preparation sit in a different category from those designed to simulate a therapeutic relationship. ThroughLine is built for the former.
ThroughLine is built around a different value entirely: agency.
ThroughLine doesn’t have a name, a personality, or a persona designed to be memorable in the relational sense. Each of the three clinical concerns above shaped specific design choices.
On the therapeutic relationship concern: ThroughLine does not position itself as a source of insight or support in its own right. It is a mirror. The AI reflects what you bring to it, asks questions to help you think more clearly, and sends that thinking toward your therapy rather than absorbing it. There is no dialogue that builds a sense of ongoing connection with the app itself.
On the safety concern: ThroughLine will not respond to distress the way a conversational AI does. It has a hard stop for crisis situations, directing users to appropriate clinical resources. It also has a usage cap, a gentle invitation to step away after a certain point, specifically designed to prevent the kind of extended interaction where conversational guardrails tend to break down.
On the false intimacy concern: the AI in ThroughLine is empathetic in manner, because acknowledgement is a useful catalyst for deeper thinking. But the acknowledgement is a mechanism, not the product. When ThroughLine recognises that something is hard, or that a feeling is present, it does so to help you think more clearly, not to make you feel understood by an app. It does not build rapport with you over time. It behaves in a particular way because it is designed and coded to. It is warm because warmth helps people think. It is not your friend.
This is a deliberate design choice grounded in a clear value. The goal of every feature in ThroughLine is to increase what you can do: what you can bring to a session, what you can hold onto between sessions, what you can notice about yourself over time. Agency, not relationship, is the point.
A therapy companion. Not a companion app.
There is a version of ThroughLine we could have built that positioned the app itself as a companion. We chose not to, and the reason is in the language we use to describe what it is. ThroughLine is a therapy companion, not a companion app. The word “companion” is doing completely different work depending on what it sits next to.
A companion app is the companion. The relationship is with the product. A therapy companion accompanies your therapy. The relationship is with your therapeutic process, and the product serves that.
ThroughLine is designed on the assumption that a therapist exists. Every feature, Reflect, Prepare, and Continue, is oriented toward that relationship: helping you bring more to it, get more from it, and sustain the momentum it generates between sessions. The emotional material you work through in ThroughLine is not meant to be resolved there. It is meant to be surfaced, clarified, and carried forward into the room with your therapist. ThroughLine is designed to send things toward therapy, not absorb them.
That is why there is no named companion. A name would imply a relationship. And the relationship that matters already has someone in it.
What must always be true. What must never be true.
We think those two questions are worth putting to any AI tool that touches therapy.
What must always be true: it should support the therapeutic relationship rather than compete with it. It should increase a client’s agency rather than their dependency. It should be transparent about what it is and what it cannot do.
What must never be true: it should never position itself as a source of therapeutic insight. It should never respond to distress in ways that fall outside a clinical framework. It should never build the kind of rapport that makes the app itself feel like a replacement for, or shortcut around, the therapeutic relationship.
That last point is the fine but distinct line ThroughLine walks. The tool needs to be somewhere people want to return to: warm enough to feel safe, structured enough to feel useful. But it is not designed to be somewhere people feel known. It doesn’t create dependency, and it isn’t designed to be used in place of therapy. It is designed so that the more you use it, the more you have to bring to the work that actually matters.
That balance is something early users noticed themselves.
“It gives you a space in between sessions. It was helpful to process things myself, but with something else.”
ThroughLine is a tool in service of something larger than itself. That is the only relationship it is trying to support.
References
Hwang et al. (2025). The defining feature of AI companionship. Emergent Mind.
Frontiers in Digital Health (2023). Your robot therapist is not your therapist. Frontiers in Digital Health.
Stanford HAI (2024). Exploring the dangers of AI in mental health care. Stanford HAI.
JMIR Mental Health (2025). A comparison of responses from human therapists and large language model-based chatbots. JMIR Mental Health.
