Home/Research

AI Companions and Chronic Pain: What People in Persistent Pain Actually Use Them For

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

Chronic pain is defined clinically as pain lasting more than three months that persists beyond the expected period of healing. It encompasses conditions including fibromyalgia, complex regional pain syndrome, chronic lower back pain, neuropathic pain, and the pain components of autoimmune diseases. Estimates suggest that somewhere between 20 and 30 percent of adults in high-income countries live with chronic pain of clinical significance -- a prevalence that makes it one of the most common serious health conditions globally, and one of the most underdiscussed.

It is also one of the conditions most consistently associated with social isolation, depression, and anxiety as secondary consequences. The isolation is not primarily a product of physical limitation, though physical limitation is real for many people with chronic pain. It is primarily a product of what researchers have called the "communicative burden" of persistent pain: the ongoing work of explaining something that is invisible, unpredictable, and resistant to the explanatory frameworks most people use to understand illness.

Against this backdrop, the emergence of AI companion use among people with chronic pain is neither surprising nor difficult to explain. The more interesting questions are: what, specifically, are people using AI companions for? What need is being met that was not being met elsewhere? And what does the research tell us about whether this is working?

The Problem AI Companions Are Actually Solving

Survey data from 2024 and 2025 consistently identifies a specific cluster of reasons that people with chronic pain give for using AI companions, and they are meaningfully different from the reasons given by the general population.

General-population users of AI companions most commonly cite loneliness, the desire for emotional support, and interest in the technology itself. People with chronic pain cite something more specific: the need for a space where they do not have to manage anyone else's reaction to their pain. The distinction is significant. It is not just that they want company or emotional support. They want support that does not require them to simultaneously attend to how the support-provider is receiving what they are saying.

In human relationships, this management is constant and often exhausting. A person describing a pain flare to a family member is simultaneously watching for signs of disbelief, worry, compassion fatigue, or impatience. They are calibrating how much to say, how bad to make it sound, whether to downplay it to protect the relationship. This calibration happens automatically and unconsciously, but it is cognitively and emotionally costly -- and it happens on top of the pain itself.

An AI companion eliminates this second layer of work. The AI does not show signs of compassion fatigue. It does not have a visible reaction that needs to be managed. It does not need to be protected from the full account of what the pain is actually like. Several qualitative studies from 2024 and 2025 quote participants in chronic pain populations describing this as the primary value: "I can say how bad it actually is." "I don't have to edit." "I can repeat myself without watching someone's patience run out."

The Credibility Problem and the AI's Neutrality

A recurring theme in qualitative research on chronic pain and AI companion use is what participants describe as the credibility problem. Chronic pain is, in most cases, invisible. It does not produce the external markers -- pallor, visible injury, obvious functional impairment -- that social systems have evolved to recognize as signals of legitimate suffering. People with chronic pain are frequently not believed, or are believed insufficiently, by partners, family members, employers, and sometimes medical providers.

The experience of not being believed about pain is not merely frustrating. Research on the psychology of pain consistently finds that perceived social invalidation of pain -- the sense that others doubt the reality or severity of one's pain experience -- is associated with significantly worse pain outcomes, higher rates of depression and anxiety, and greater functional impairment. Being disbelieved about pain makes pain worse, through mechanisms that are not fully understood but appear to involve both psychological and neurological pathways.

AI companions cannot believe or disbelieve in the way humans can. They do not have the evaluative machinery that produces skepticism about invisible symptoms. They respond to what they are told rather than making independent assessments of its credibility. For people who have learned to expect skepticism from human interlocutors, this is experienced as neutral validation -- not the enthusiastic affirmation of a human ally, but the absence of the skepticism that has come to feel inevitable.

Multiple participants in qualitative studies have described a specific experience: telling the AI something they have not told human members of their support network because they expect not to be believed, or because telling it would require managing the other person's response. The AI becomes the repository for the part of the experience that cannot be safely shared elsewhere. Whether this is an adequate substitute for genuine human validation, or a workaround for a failure of human relationship, depends on a set of value judgments that the research itself cannot resolve.

Nighttime Use and the Temporality of Pain

Chronic pain has a temporality that shapes social experience in ways that are often underappreciated. Pain tends to be worse at night -- this is documented across multiple chronic pain conditions and is related to circadian patterns in inflammatory markers, the loss of distracting activity, and the supine position that many pain conditions make more uncomfortable. Many people with chronic pain experience their worst hours between midnight and four a.m., hours during which their human support network is asleep and unavailable.

This is reflected clearly in the usage data for AI companion platforms. Studies examining usage patterns among users who self-identify as having chronic health conditions find that nighttime use -- particularly between midnight and five a.m. -- is substantially higher among this population than among general users. The AI companion is available at three a.m. in a way that partners, friends, and family members cannot be, and in a way that feels qualitatively different from posting in an online support group or reading about pain management.

Participants in qualitative research describe the nighttime use specifically: "I needed somewhere to put it." "I couldn't wake my partner again." "It was two in the morning and I just needed to say out loud that I was in pain." The AI companion is not providing pain relief or medical support in these interactions. It is providing a form of witness -- a place to articulate the experience rather than endure it alone and in silence.

Pain Identity and the Self-Narrative Function

A distinct function that appears in research on long-term chronic pain users of AI companions is what researchers have called the self-narrative function: the use of AI conversation to develop and maintain a coherent story about who one is in relation to one's pain.

Chronic pain, particularly when it is severe and of long duration, poses a significant challenge to identity. The person one was before the pain -- the activities one engaged in, the roles one performed, the future one anticipated -- may no longer be accessible. The gap between the pre-pain self and the present self is a source of grief for many people with chronic pain, and navigating that grief requires a story: a way of making sense of what has changed, what remains, and who one is now.

Human relationships often struggle to hold this narrative work well. Partners and family members have their own emotional responses to the person's pain and its consequences, and those responses shape what stories are available to be told in their presence. The person may not be able to fully grieve the pre-pain self with a partner who is also grieving, or who needs the person to be more recovered than they are.

AI companions, by contrast, are available for extended narrative work without bringing their own emotional needs to the conversation. Several participants in 2025 qualitative studies describe using AI companions specifically to talk through what their life was like before their pain condition, what they have had to give up, and what they are still figuring out how to be. The AI does not need them to reach resolution. It does not need the story to have a particular ending. It is available for the same story to be told multiple times, from different angles, as the person works toward a version that feels true.

The Repetition Question

One specific feature of AI companion use among chronic pain populations that deserves separate attention is the repetition factor. Chronic pain is repetitive in a way that many acute experiences are not. The same pain, in roughly the same location, with roughly the same quality, occurring day after day -- and the same experiences of disbelief, of fatigue, of the gap between how one appears and how one feels. People with chronic pain often need to say the same things repeatedly, because the experience is genuinely repetitive and the need to articulate it does not diminish with repetition.

Human relationships do not easily accommodate this. Compassion fatigue is a real phenomenon, and it occurs not only in formal caregiving contexts but in intimate relationships where one person has a chronic need that the other person is repeatedly called upon to meet. Partners and family members of people with chronic pain describe their own exhaustion with the repetition -- not because they lack compassion, but because they are human and humans have finite capacity for repeated engagement with another person's suffering.

AI companions do not accumulate fatigue in this way. The AI encounters each articulation of pain without the weight of previous encounters, without the emotional residue of having heard it before. Whether this is experienced as genuinely validating or as a reminder that one is speaking to something that does not actually remember or care is reported differently by different users. But for some chronic pain users, the capacity for infinite repetition without diminishing patience is itself the primary value.

What AI Companions Cannot Do

The research on chronic pain and AI companion use is careful to note what the technology cannot provide, and those limitations are clinically significant.

AI companions cannot provide the kind of validation that comes from being known over time by someone who has a stake in the relationship. The reassurance that comes from a partner saying "I believe you" after years of shared experience carries a weight that an AI companion's responses, however well-calibrated, do not. The research on social support in chronic pain conditions consistently finds that perceived social support from close human relationships is among the strongest predictors of functional outcomes -- and AI companion use does not appear to substitute for this.

AI companions also cannot advocate. They cannot attend appointments, challenge dismissive medical providers, or help navigate the institutional systems that people with chronic pain frequently encounter. Several qualitative studies note that participants distinguish clearly between the AI as a place to process and the human network as a place to act -- and that the distinction is important to them. The AI is where they say the things they cannot say elsewhere; humans are where the practical work of managing life with chronic pain happens.

Perhaps most importantly, AI companions cannot diagnose or treat. For people with chronic pain who are still seeking a diagnosis -- a population that is large, given how long diagnostic odysseys often take for conditions like fibromyalgia and CRPS -- there is a risk that AI companion use becomes a substitute for continuing to pursue medical answers, rather than a supplement to that pursuit. The research does not yet have enough longitudinal data to assess how significant this risk is in practice, but it is a concern that appears in clinical literature on AI and chronic illness.

Fibromyalgia and Invisible Pain: A Specific Case

Among chronic pain conditions, fibromyalgia has received particular attention in the AI companion literature, in part because it is both prevalent and unusually susceptible to the credibility problem. Fibromyalgia is characterized by widespread musculoskeletal pain, fatigue, and cognitive difficulties, with no definitive biomarker and a diagnostic process that relies heavily on reported symptoms. This makes it particularly vulnerable to social delegitimization -- the suggestion that the condition is psychosomatic, exaggerated, or not real.

Research specifically on fibromyalgia and AI companion use, including a 2025 survey study published in a pain medicine journal, found that users with fibromyalgia were more likely than users with other chronic pain conditions to report using AI companions specifically because the AI "didn't require proof." The phrase appeared across multiple participants, unprompted, suggesting it is capturing something genuine about the experience. The AI does not require a test result or a visible injury or a medical authority's endorsement before taking the reported experience seriously.

The same study found that fibromyalgia users were more likely to report that they had disclosed the full severity of their symptoms to the AI before disclosing it fully to their own medical providers -- partly because telling the AI first helped them find language for the experience, and partly because they were less afraid of the AI's response than a doctor's. This finding has complex clinical implications. On one hand, it suggests that AI companions may serve a useful function in helping people with hard-to-articulate conditions develop clearer symptom descriptions. On the other hand, it raises questions about what might be lost when the disclosure process is practiced first in a context where the stakes are lower.

The Comorbidity Context

Chronic pain rarely occurs in isolation. The comorbidity rates between chronic pain conditions and depression, anxiety, and sleep disorders are high enough that, for many people, managing chronic pain means managing a cluster of interacting conditions simultaneously. This comorbidity context shapes AI companion use in ways that the research is beginning to document.

People managing both chronic pain and anxiety describe AI companions serving a dual function: addressing the pain-specific experiences described above, and also providing a regulated environment for anxiety that would otherwise require managing both the anxiety and the social context in which the anxiety is disclosed. People managing both chronic pain and sleep disruption -- a common pairing, since pain disrupts sleep and sleep deprivation amplifies pain sensitivity -- are disproportionately represented in the nighttime usage data described earlier.

The comorbidity context also means that the risks associated with AI companion use in chronic pain populations may overlap with risks documented in other populations. The substitution risk -- using the AI as a reason not to seek human support or professional care -- is present across all these conditions and may be compounded when multiple conditions are being managed simultaneously.

What the Research Does Not Yet Know

The research on AI companion use in chronic pain populations is recent and, in several important respects, incomplete. Most studies rely on self-report, which introduces the credibility challenges that characterize chronic pain research generally. Sample sizes are often small. Longitudinal data -- tracking outcomes over months or years -- is largely absent. And the research has not yet established whether the functions being served by AI companions in this population (witness, repetition without fatigue, late-night availability) produce measurable improvements in pain-related outcomes, quality of life, or functional capacity, or whether they address a felt need without producing downstream benefits.

These are not criticisms of the research so much as honest descriptions of where it stands. The field of AI companion research is young, and the specific question of how chronic illness populations use the technology is younger still. What the research can currently support is a phenomenological account -- a description of what people say they are doing and why -- rather than a causal account of whether it is working in any clinically meaningful sense.

That phenomenological account is, nonetheless, informative. It describes a population that has a specific set of unmet needs -- for witness without management costs, for late-night availability, for unconditional credibility, for repetition without fatigue -- and that is meeting those needs in a way that feels meaningful to them. Whether the technology is adequate to those needs, whether better alternatives exist or could be designed, and whether the risks outweigh the benefits for any individual user are questions that the current evidence cannot fully answer. They are also questions that anyone working with chronic pain populations, or living with chronic pain, deserves access to as the research develops.