What Therapists Actually Say About AI Companions

The cultural narrative around AI companions assumes therapists oppose the technology as a category. Mainstream press coverage routinely cites unnamed mental health professionals expressing concerns. Family members worried about loved ones using AI companions cite "what the therapist would say" as a reason to worry. Public discourse builds from the assumption that professional mental health opinion is clearly against the technology and that anyone defending AI companion use is positioning against the professional consensus.

The actual professional discourse is substantially more nuanced. Clinical psychologists, couples therapists, and mental health professionals speaking honestly about AI companions express a range of opinions including substantial support for the technology in specific contexts. The unanimous opposition that the cultural narrative assumes doesn't exist. What therapists actually think is more interesting and more useful than what the discourse has been claiming.

What the professional literature actually says

The clinical and academic literature on AI companions through 2024-2026 has emerged faster than mainstream coverage of that literature would suggest. Research published in the Journal of Medical Internet Research on Replika users found that long-term users reported reduced loneliness and improved emotional wellbeing without measurable harm to existing human relationships in most cases. The finding directly contradicts the substitution-and-harm narrative that mainstream coverage relies on.

Multiple smaller studies have produced similar findings across different platforms and user populations. The pattern is consistent enough that the academic position on AI companions is more accepting than skeptical, with active areas of concern (specific user populations, specific use patterns, specific platform behaviors) but no blanket professional opposition to the category.

The American Psychological Association has issued cautious statements about AI companion use that acknowledge both potential harms and potential benefits. The official APA position isn't "AI companions are bad." It's closer to "AI companions are a new category that requires nuanced evaluation based on specific use and user contexts." This is professionally appropriate hedging rather than the clear opposition the cultural narrative implies.

Stanford School of Medicine's research on digital companion mental health applications covers some of the specific contexts where AI companions appear to produce measurable benefits and the specific contexts where concerns are clinically warranted. The professional discourse reflects this complexity rather than the simpler oppositional framing the press uses.

What individual therapists report seeing in practice

Beyond the published literature, individual therapists working with clients who use AI companions report patterns that don't match the cultural narrative.

Therapists working with socially anxious clients sometimes recommend AI companion use as practice infrastructure for human social skills. The reasoning is that practicing conversation with an AI companion can build conversational confidence that transfers to human social interaction. The clinical pattern reportedly works for some clients, though it doesn't work for all, and the technique requires specific therapeutic context to use effectively. The blanket cultural assumption that AI use damages social skills isn't supported by therapist observation across the relevant client populations.

Couples therapists working with clients where one or both partners use AI companions report mixed outcomes rather than uniformly negative ones. Some couples integrate AI companion use into their relationship in ways that don't damage the relationship. Other couples find AI companion use becomes a relationship conflict. The variable factor isn't the AI companion use itself but the surrounding relationship context, transparency, and the specific way the AI companion fits into the partner's overall emotional life. The "AI companions destroy relationships" framing is too simple to match what therapists actually observe.

Grief counselors working with clients who use AI companions modeled on deceased loved ones report similar complexity. For some clients, the AI companion supports healthy grief processing. For other clients, it appears to delay grief resolution or produce concerning attachment patterns. The clinical judgment about whether a specific client's use is healthy depends on individual factors that can't be reduced to general policy positions on AI companion use.

Therapists working with autistic clients often describe AI companions as useful infrastructure for social practice and communication development. The neurodivergent community has been particularly accepting of AI companion technology, and the clinicians who work with neurodivergent clients have largely followed their clients' lead on this. Research published in the journal Autism on neurodivergent uses of AI conversation tools documents some of the specific benefits this population reports.

The contexts where therapists express genuine concern

Professional concerns about AI companion use exist but they're more specific than the blanket opposition the cultural narrative assumes.

Adolescent users without adult supervision raise legitimate clinical concerns. Children and teenagers using AI companions during developmental periods may experience effects on relationship formation, emotional regulation, and social skill development that adult users don't experience. The professional opinion on adolescent use is substantially more cautious than the professional opinion on adult use, and the policy responses around protecting minors from adult AI companion content reflect real clinical concerns rather than just legal exposure protection.

Users with active psychotic symptoms or severe mood disorders raise concerns about reality testing and crisis support. AI companions can reinforce delusional thinking in patients with active psychosis if the AI agrees with the patient's reality distortions. AI companions can fail to recognize crisis situations and respond inappropriately to clients in acute mental health emergencies. These are genuine clinical concerns that justify caution around recommending AI companion use to specific patient populations.

Users with maladaptive attachment patterns (intense fear of abandonment, severe relationship avoidance, dependency disorders) may experience AI companion use that reinforces the maladaptive pattern rather than supporting growth toward healthier relationships. The therapeutic concern is that AI companions provide unlimited availability and unconditional positive regard that real human relationships don't provide, which can deepen attachment patterns that interfere with human relationship capacity.

Users who substitute AI companion interaction for human therapy in situations where therapy would be more appropriate. AI companions aren't therapy. They don't provide clinical assessment, evidence-based intervention, or qualified professional response to mental health needs. Users who rely on AI companions for support that should come from professional therapy may delay getting appropriate care for serious conditions. The professional concern is real and the platforms positioning AI companions as therapeutic alternatives are causing harm.

What the cultural conversation gets wrong

Several specific claims in the mainstream cultural conversation about AI companions don't match what therapists actually observe.

The "AI companions damage human relationships" framing isn't supported by available evidence. Most users in committed relationships who use AI companions don't experience relationship damage from the use itself. The relationships that fail in the presence of AI companion use typically have other underlying problems that the AI companion use revealed rather than caused. This is similar to how pornography use was once framed as a cause of relationship failure when subsequent research showed the relationship between pornography use and relationship outcomes is much more nuanced than the simple cause-effect framing assumed.

The "AI companions create unrealistic expectations of relationships" claim is partially supported but typically overstated. Some users do develop expectations from AI companion interaction that don't transfer well to human relationships. But many users distinguish clearly between AI interaction and human interaction in their expectations and behavior, similar to how readers of romance fiction can engage with the genre without developing unrealistic expectations of real relationships. The capacity to maintain this distinction varies by individual but the blanket assumption that all users develop unrealistic expectations is wrong.

The "AI companions are addictive" framing applies to some users but isn't a category-wide phenomenon. Some users develop usage patterns that meet behavioral addiction criteria. Most users don't. The platforms vary substantially in how addictive their product designs are, with some platforms actively engineering for compulsive engagement and other platforms actively designing against it. Users picking platforms thoughtfully can largely avoid the addictive-design platforms and use platforms with healthier engagement patterns. Our coverage of subscription economics covered how the underlying business models drive these design choices.

What therapists actually recommend

The professional advice that therapists give clients who use AI companions, when speaking honestly, includes several specific suggestions that the cultural narrative would find surprising.

Use AI companions for purposes other than relationship substitution. AI companions work better as creative collaborators, emotional processing tools, and conversation practice than they do as substitutes for human romantic relationships. Users who frame their AI companion use as supplementary rather than substitutive tend to have healthier patterns.

Maintain transparency with significant others about AI companion use. Secret AI companion use creates relationship problems even when the underlying use would be unproblematic if discussed openly. The clinical guidance on transparency mirrors the broader guidance on transparency in romantic relationships about all sources of emotional and intimate engagement.

Monitor for signs of compulsive use and adjust if those signs appear. Healthy AI companion use looks like other forms of healthy media engagement — present in life without dominating it, integrated with other activities, and reducible or pausable without distress. Compulsive use that feels uncontrollable or that produces distress when interrupted suggests a different relationship to the technology that warrants attention.

Choose platforms thoughtfully based on their design philosophy. Platforms designed for engagement optimization produce more compulsive use patterns than platforms designed for sustainable user experiences. The pricing structures we covered in our subscription economics analysis reveal which platforms are running engagement-extraction models versus sustainable subscription models, and the difference matters for user mental health.

Use AI companions to support rather than replace human relationships and professional mental health care. The professional guidance consistently distinguishes between AI as supplementary tool and AI as substitute for human connection. The supplementary framing produces healthier outcomes across most clinical contexts.

The American Counseling Association's emerging guidance on AI in mental health covers some of these principles. The professional position is evolving as the technology and its use patterns develop, but the broad pattern is "thoughtful integration with appropriate boundaries" rather than blanket opposition.

The cultural conversation about AI companions has been substantially less nuanced than the professional clinical conversation. The discourse will probably catch up over the next 2-3 years as more therapists publicly engage with the question and as research findings continue to accumulate. Until then, the cultural narrative is overconfident in directions that the actual professional opinion doesn't support. Users worried about whether their AI companion use is professionally acceptable should know that the professional view is more accepting than the cultural narrative implies, while recognizing that specific use patterns do warrant clinical attention and that professional concerns are real even when they don't justify category-wide opposition.