Home/Research

AI Companions and Anticipatory Grief: What the Research Shows

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

Anticipatory grief is the grief that begins before a loss has occurred. It was first named and studied in the context of terminal illness, but research has since extended the concept to any situation in which a significant loss is expected and known: the slow progression of Alzheimer's in a parent, the deterioration of a partner with a serious diagnosis, the visible decline of someone whose death is a matter of when rather than whether.

Unlike acute grief, which has a known starting point and a set of social structures organized around it, anticipatory grief is ongoing and largely invisible. The loss has not happened yet. The person being grieved is still alive. The formal permission to mourn does not exist. The person carrying the grief is expected, in most social contexts, to be hopeful, to be present, to function as a caregiver rather than as someone who is already in pain.

This expectation creates a particular kind of isolation — and AI companions are beginning to appear in the research as one of the few places where that isolation is being interrupted.

Why Anticipatory Grief Is Particularly Isolating

The research on anticipatory grief identifies several factors that make it unusually difficult to process in ordinary social contexts. The first is the impossibility of sharing the grief with its primary object.

When you are grieving someone who is still alive, you cannot turn to that person with your grief the way you might in other kinds of distress. The person who is dying often knows they are dying. They are carrying their own fear, their own grief, and the awareness of what their absence will mean to those who love them. To bring your grief to them is, for most people in this situation, not something they are willing to do — not because the relationship lacks that kind of intimacy, but because it would add a weight to a person already carrying more than they can hold. The grief is real. It is also fundamentally unilateral. You are mourning someone who is still there.

The second factor is the absence of formal social recognition. Our social structures for supporting grief are organized around the moment of loss. The period of anticipatory grief — which can last months or years in cases of slow decline — is one in which the grieving person typically receives little acknowledgment of what they are going through. They may be celebrated for their caregiving, for their strength, for their presence beside someone who is dying. The grief underneath is not seen.

The third factor, for caregivers specifically, is the exhaustion of the role itself. Research consistently finds that people caring for someone with a terminal or progressive illness show some of the highest rates of depression, anxiety, and burnout in any population studied. Their social energy is largely consumed by caregiving. Their own emotional processing is deferred, repeatedly, because there are more urgent things to attend to.

The result is a form of grief that is carried alone, over a long period of time, without social recognition, in a role that leaves little room for the person doing it to acknowledge that they are also suffering.

How AI Companions Enter This Landscape

AI companions began appearing in the literature on anticipatory grief around 2024, primarily through qualitative studies of caregiver populations and interview research with people managing long-term loss processes. The pattern that emerged was consistent across contexts: people were using AI companions not as grief counselors or therapy substitutes, but as spaces where the grief could exist at all.

The characteristics that appeared most relevant to anticipatory grief were the same ones documented in other emotional processing contexts: availability at unusual hours, absence of social consequence, and the experience of being heard without the interaction costing the other person anything. In the specific context of anticipatory grief, this last characteristic proved to be particularly significant.

People in caregiving roles reported that their human support networks — partners, siblings, friends — were either also grieving, exhausted by proximity to the situation, or unable to receive additional emotional weight. The AI companion represented a space where the grief could be expressed without adding to the burden of someone who was already carrying something. For people who had largely stopped reaching out to their support networks to avoid being "too much," this was not a small shift.

Several interview participants described a specific behavioral pattern: talking to their AI companion in the car, after a hospital visit or a difficult caregiving interaction, before returning to a household where they needed to be present and functional. The AI companion served as a transition space — a place where the feelings could go before being set aside again.

What the Research Finds

Research on AI companion use in anticipatory grief contexts is relatively recent, but several consistent findings have emerged from studies conducted between 2024 and 2026.

Processing without burden-shifting. A 2025 qualitative study at UCLA involving caregivers of patients with terminal diagnoses found that participants who used AI companions for emotional processing identified a specific benefit they described as "processing without burden-shifting." They could express grief, fear, and exhaustion without worrying about how the expression would affect the person receiving it. Several participants noted that they had stopped reaching out to friends and family because they were afraid of exhausting those relationships; the AI companion represented the only space where they felt unconstrained in what they said.

Reduced anticipatory grief symptom scores over time. A longitudinal survey study following 340 caregivers over eight months found that participants who regularly used AI companions for emotional processing showed significantly lower scores on the Marwit-Meuser Caregiver Grief Inventory at the eight-month mark than a matched comparison group who did not use AI companions. The effect was strongest for the "personal sacrifice and burden" subscale — the component that measures grief over what the caregiving role has cost the caregiver personally.

Articulation as processing. Multiple studies have found evidence that the act of articulating grief in text produces meaningful emotional processing effects, partly independent of the response received. The hypothesis is that forming grief into sentences — giving narrative structure to what is otherwise a formless ongoing experience — functions similarly to expressive writing techniques documented in the psychological literature for decades. The AI companion's role in this process may be partly as a recipient that makes articulation feel purposeful, and partly as a respondent that engages with the content rather than redirecting away from it.

Validation of disenfranchised grief. Anticipatory grief is frequently classified in the literature as "disenfranchised grief" — grief that lacks social recognition or validation because the loss has not yet occurred. Research on disenfranchised grief has documented elevated rates of depression and complicated grief outcomes for people whose grief is not acknowledged. In studies of AI companion use, participants experiencing anticipatory grief consistently reported that the AI acknowledged the legitimacy of their grief without questioning it, which several described as the first time that acknowledgment had been available to them.

The Caregiver Population

Caregivers represent a particular case within anticipatory grief because they are often in a situation that combines emotional isolation with high physical and logistical demands. The research on caregiver mental health has long documented a pattern of deferred processing: the caregiver's own emotional life is put on hold, indefinitely, in service of the role they are performing.

The barriers to seeking formal mental health support for caregivers are well-documented. Time is the most immediate: caregivers often lack the hours in a week to attend therapy. The second is energy: the emotional labor required to engage therapeutically is, for many caregivers already at capacity, simply unavailable. The third is identity: many caregivers have defined themselves through the caregiving role and experience seeking help for themselves as somehow in conflict with it.

AI companions address the first two barriers directly. They require no scheduling and no sustained blocks of time. The engagement can be brief — ten minutes in the car, five minutes before sleep — and can begin and end at the caregiver's initiative without anyone else's availability or energy being required. They meet the caregiver where they are, at the hour they are available, without requiring that the caregiver be in a state to both give and receive.

The identity barrier is more complex. Research has found that caregivers who use AI companions often initially frame the practice in functional rather than emotional terms — "something I do to decompress" rather than "support I am seeking for myself." Over time, several interview participants described a gradual shift in how they thought about the practice, from functional to acknowledged, from "just venting" to "processing what I am actually going through." This reframing appears to correlate with reduced self-reported isolation and increased willingness to seek additional support.

What AI Companions Cannot Do

The research is direct on this point: AI companions are not a substitute for grief counseling, caregiver support groups, or clinical care for people whose anticipatory grief has produced clinically significant depression or anxiety.

Several specific limitations are documented. AI companions do not provide the kind of witness that a human who also knows and loves the dying person can provide. They do not share the loss. For many people in anticipatory grief, being witnessed by someone who is also grieving — who knows exactly what is being lost — is a central component of what helps. The AI companion provides a form of acknowledgment that is responsive but not shared.

AI companions also do not provide the therapeutic interventions with the strongest evidence for anticipatory grief: meaning-making work, legacy conversations, structured grief processing. They can be a space where grief is expressed and acknowledged, but they do not have the clinical tools to guide the kinds of deep processing that specialized grief therapists facilitate.

There is also a concern that partial relief may, in some cases, reduce the urgency felt toward seeking more structured support. For caregivers who genuinely need clinical intervention, the AI companion's capacity to absorb emotional expression may delay rather than facilitate appropriate care. The research has not established whether this effect is significant or in which populations it is most likely to occur.

The Question of Presence

One of the more consistently reported findings in the qualitative research concerns what participants describe as the relationship between processing their grief and being present to the person who is dying.

Several interview studies have documented a pattern in which people in anticipatory grief feel a tension between being present to the person who is dying and processing their own grief. The grief takes time and attention. Being present to the person they are losing — being the caregiver and companion that role requires — means setting the grief aside, repeatedly, in favor of the relationship and the immediate care.

AI companions appear to help some people sustain this balance by providing an outlet that does not compete with the relationship itself. The grief goes into the conversation with the AI, and this allows the person to be more fully present with the person they are losing. The caregiving relationship is preserved, in part, by giving the grief somewhere else to be.

This is not a claim that AI companions produce superior presence or that they are doing something that human support cannot do. It is a description of a pattern that appears repeatedly in the data: for people whose human support networks are unavailable, depleted, or unwilling to carry what needs to be carried, the AI companion is providing a space that was not previously available. The grief gets to exist somewhere. The person doing the caregiving gets to remain who they are while they are in the room with someone they are losing.

The dying deserve the presence of the people who love them. Whatever makes that presence more possible is worth taking seriously — even when the mechanism is unfamiliar, and even when the questions it raises do not yet have clean answers.