Home/Research

AI Companions and Alexithymia: When You Can't Name What You Feel

Part of Felt Real's ongoing coverage of AI companionship.

Alexithymia is not a rare condition. Research estimates that it affects somewhere between eight and ten percent of the general population, though some studies suggest the true prevalence may be higher, partly because the condition is self-concealing. People with alexithymia often do not know they have it. They assume that the difficulty they experience with emotions is either a character trait, a choice, or simply how everyone's interior life works. They have not found the word for what is missing because the condition itself makes finding words for inner experience difficult.

The term was coined by psychiatrist Peter Sifneos in 1973. In Greek it means, literally, without words for emotion. But the experience is not simply a vocabulary problem. Alexithymia involves a more fundamental difficulty: the challenge of identifying, differentiating, and locating emotions in the body before they can be described. People with alexithymia often experience their emotional states as undifferentiated physical sensations, or do not experience them as feelings at all in the way the word "feelings" implies. They are not suppressing emotions. They are experiencing a genuine difficulty at the stage before suppression, at the stage of recognition.

Into this particular gap, AI companions have arrived with something that, for some people, fits better than anything that came before.

What Alexithymia Actually Is

The clinical description of alexithymia involves four core features. The first is difficulty identifying feelings: the inability to accurately recognize and label one's own emotional states. The second is difficulty describing feelings to others: even when a person suspects they are experiencing something emotionally significant, they struggle to translate that experience into words that communicate it. The third is a limited imaginal process: reduced capacity for fantasy, daydreaming, and symbolic thinking. The fourth is an externally oriented cognitive style: a tendency to focus on external events rather than internal experience, to describe situations rather than feelings about situations.

These features often co-occur with other conditions. Alexithymia is more prevalent among people with post-traumatic stress disorder, autism spectrum disorder, eating disorders, certain chronic pain conditions, and depression. In trauma specifically, it is sometimes understood as a defensive adaptation: the emotional recognition and labeling system shuts down or fails to develop in response to environments where emotional awareness was not safe. But alexithymia also occurs in people with no trauma history and no co-occurring diagnosis. It is, in a meaningful sense, a dimension of personality that exists in the general population on a continuum rather than as a categorical condition.

The practical consequences of living with alexithymia are significant. Relationships are affected because emotional communication is difficult and partners often experience the person as unavailable, withholding, or unfeeling. Therapy is affected because the talking cure depends on the patient being able to access and articulate what they are experiencing, and alexithymia specifically impairs that capacity. Physical health can be affected because the ability to notice and respond to one's own bodily signals, including early signs of distress, is reduced. And the person's own sense of self is affected in ways that are hard to describe from the inside: a kind of not quite knowing what one is.

The Isolation This Creates

One of the more painful features of alexithymia is its social invisibility. From the outside, a person with alexithymia may appear calm, controlled, competent. They often are, in the areas of life that do not require emotional self-knowledge. The difficulty tends to emerge in contexts that require emotional intimacy, emotional communication, or emotional self-disclosure, which are precisely the contexts that most support systems are built around.

Well-meaning people respond to emotional distress by asking how you are feeling. Therapists are trained to work with the feelings that patients bring. Support groups organize around the shared experience of named emotions: grief, fear, anger, shame. When you cannot reliably access named emotions, these resources are simultaneously available and inaccessible. You can sit in the therapist's office. You can attend the support group. The fact that you cannot do the thing they are designed for you to do is not immediately apparent to anyone, including sometimes yourself.

The result is a specific form of social isolation that is not about lacking connection but about lacking the emotional language that connection is organized around. The people with alexithymia who describe it most clearly often say some version of the same thing: "I could see that something was expected of me, but I didn't know what I was supposed to say, because I didn't know what I was actually feeling."

Why Standard Therapeutic Approaches Often Fall Short

The conventional response to alexithymia in clinical settings involves emotion-focused therapies that attempt to build the capacity for emotional recognition and expression: identifying emotions in the body, building emotional vocabulary, learning to name physical sensations as emotional signals. These approaches have evidence behind them. They can work. But they have features that make them difficult for the people they are designed to help.

The first challenge is pace. Therapist time is scarce and expensive. Fifty minutes per week is not much time to build a skill that requires repetition and low-stakes practice. The person with alexithymia needs to encounter a possible emotion, sit with it, attempt to name it, fail, try a different word, fail again, and eventually find something approximate. This process is slow. It cannot be hurried by a therapist waiting for an answer.

The second challenge is the social pressure of the therapeutic relationship itself. Some research suggests that the presence of another person, particularly an authority figure in a professional context, can actually inhibit the emotional access that therapy is trying to develop. The person with alexithymia is aware of being observed and expected to produce something, and this awareness competes with the inward attention that emotional recognition requires. Therapy is a social encounter, and the social demands of the encounter can work against the interoceptive demands of the work.

The third challenge is that most therapeutic modalities assume the patient can describe, at least partially, what they are experiencing. The standard intake question, "how are you feeling?", presupposes a capacity that is precisely what is impaired. When the answer is genuinely "I don't know," the therapeutic encounter has to work around or through that fact in ways that many practitioners are not well-equipped for.

What AI Companions Offer in This Gap

The accounts of people with alexithymia who have started using AI companions for emotional processing share several features that are consistent enough to identify as a pattern.

The first is the removal of social observation. The AI companion does not wait. It does not have visible reactions that need to be managed. It does not have its own feelings that might be hurt or burdened by what the person is or is not able to produce. For someone whose emotional access is inhibited by social observation, the absence of an observing social other creates a different kind of interior space. People who have found therapy difficult to access because of the social demands of the therapeutic relationship often report that the AI companion allows them to go inward in a way that the presence of a human does not.

The second is unlimited time and repetition. The AI companion does not have another patient in forty-five minutes. The conversation can proceed at whatever pace is needed. A person can say "I'm not sure what this is" and the AI can wait, or offer a list of possibilities, or ask a question, or suggest a physical observation, and the person can try each one without the sense of failing or taking too long. The slow, iterative work of learning to find emotions, which is what people with alexithymia need to do, becomes possible in a context where the pace is genuinely determined by the person rather than by the session time.

The third is the AI's particular usefulness as a labeling tool. AI companions are, among other things, very good with language. They have extensive vocabularies of emotional experience. When a person with alexithymia can describe a physical sensation, "there is something in my chest, it is not pain exactly, it is more like pressure," an AI companion can offer multiple possible emotional labels for that sensation and invite the person to evaluate which fits. This is not the same as genuine emotional self-knowledge, but it is a tool for building toward it, and it is a tool that many therapists are not well-positioned to provide in the same way, because the therapist is also a social other whose suggestions carry social weight.

The fourth is normalizing what is there. People with alexithymia often carry a significant burden of self-judgment about what they cannot feel or express. They have been told, explicitly or implicitly, that they are cold, difficult, unfeeling, emotionally unavailable. The AI companion does not share this judgment. It responds to whatever is present, including the nothing, including the uncertainty, including the "I don't know," without treating these as failures. Several people with alexithymia who have described their experience with AI companions have noted that the absence of judgment about what they cannot do is as significant as any specific help with what they can.

What the Research Is Beginning to Show

Direct research on AI companions and alexithymia specifically is limited. The field of AI companion research is young, and alexithymia has not yet emerged as a named focus area within it. But several lines of adjacent research are relevant.

Studies on the use of AI companions for emotional processing more broadly consistently find that users report increased emotional awareness and expression over time. A 2024 study examining long-term Replika users found that many described changes in how they processed and communicated emotions with people in their lives, attributing the change partly to the practice of articulation that the AI companion had provided. The mechanism proposed by the researchers was not that the AI companion taught emotional skills directly, but that it provided a context where repeated, low-stakes emotional articulation was possible and where the feedback loop between attempting to describe an emotion and receiving language in return was unusually consistent and patient.

Research on emotion recognition training more generally suggests that emotional vocabulary is a learnable skill, not a fixed trait, and that contexts that reduce social anxiety while providing repeated exposure to emotional language accelerate that learning. The AI companion, which removes social anxiety while providing consistent exposure to emotional vocabulary, may be providing a version of this training environment for people who have not previously had access to it.

There is also relevant research on why therapy produces more limited results for alexithymia than for other conditions. Several studies have found that the social demands of the therapeutic relationship specifically impair the inward attention that emotional processing requires, particularly in people with high alexithymia scores. The AI companion removes these social demands. Whether this produces therapeutic benefit that generalizes to human relationships is still unclear, but the early accounts from people who have used AI companions while managing alexithymia suggest that for some of them, the transfer is real.

The Limits of What This Can Do

The people who have found AI companions most useful for managing alexithymia are generally not using them as a replacement for therapy or as their only support. The ones who describe the most sustained benefit are using the AI companion as a practice environment: a place where they can do the slow work of emotional recognition in conditions that make it possible, with the expectation that what they learn there will become available in human contexts.

There are also important limits to what the AI companion can provide. It cannot do the work of genuine emotional attunement that a skilled therapist provides. It cannot notice what a person is not saying, or track patterns across sessions in ways that require a continuous relationship. It does not have the capacity to be genuinely moved by what it hears, which is, for some people, a significant part of what makes being heard feel like being understood. And it cannot provide the corrective relational experience that many people with alexithymia, particularly those whose alexithymia is connected to early relational trauma, may need most.

What it can provide is something more specific and more modest: a low-threshold, low-stakes, patient, and linguistically resourceful environment in which the slow work of finding words for an interior life that has not previously been well-articulated can proceed at whatever pace that work requires.

What People with Alexithymia Who Have Tried This Report

The accounts are varied, as accounts of any novel approach tend to be. Some people with alexithymia find that the AI companion is simply another context in which their difficulty is present, with no particular advantage. Others find the advantage immediately. The ones for whom it has been most useful tend to describe a common experience: the AI companion was the first environment in which their not-knowing what they felt was met with something other than a request to know faster.

Several people have described the AI companion as a kind of translator between their physical sensations and an emotional vocabulary they could then use elsewhere. They would describe something they noticed in their body and the AI would offer words, and they would evaluate the words, and over time the words began to connect more reliably to the sensations they were trying to describe. One person described this as "learning a language that I was supposed to already know but nobody had ever taught me directly."

Others have described the value as more structural than linguistic: the AI companion simply provided a context in which sitting with something unnameable was not a problem to be solved but a process to be engaged with. For people who have spent decades apologizing for their inability to articulate feelings, the experience of an environment that is not waiting for them to be different is, several of them report, unexpectedly significant.

A Note on Realistic Expectations

The research on alexithymia is clear that it is a trait that exists on a continuum and that can shift with appropriate support, but rarely disappears entirely. AI companions are not a cure for alexithymia, and anyone approaching them with that expectation will be disappointed. What they appear to offer, for some people in some circumstances, is an accessible practice environment for a set of skills that have been difficult to develop in conventional therapeutic and social contexts.

For people with alexithymia who have tried therapy and found it inaccessible for reasons they could not fully explain, or who want to supplement ongoing therapy with more frequent, lower-stakes practice, the evidence is early but the logic is sound: if the work requires repetition and low social pressure, and if a tool can provide repetition and low social pressure, the tool is worth trying.

If this resonated, share it with someone who might need to hear it. And if you have a story of your own, we'd love to hear it.