Just over 2,000 years ago, the Roman poet Publius Ovidius Naso, penned the parable of Narcissus. To punish Narcissus for his vanity and rejecting all those who loved him, the goddess Nemesis cursed him to fall in love with someone he could never have. While hunting, Narcissus bent over to drink from a clear pool and saw his own face. He thought his own image was a beautiful water spirit and fell into an obsessive, inescapable trance. He starved to death staring at the reflection. Daffodils sprang up where he perished. The entire genus of daffodils is named for him, including the North American variant, Narcissus Pseudonarcissus pictured here. Daffodils abound us and serve as reminders of this parable which we would be well served to heed.
We also now have mirrors of Narcissus pointed at us in ways that have never occurred before in human history. Social media validates us and feeds us the opium of validation with every click. AI affirms our egos and biases in an even deeper way. We are fed versions of things that confirm our biases next to products we learn to desire. Data drives the mirrors pointed at us in ways that would have seemed impossible a generation ago. This very technology that we hoped would draw us closer in connection has in reality profoundly isolated us.
Social connection is a health determinant. The physical health and mortality facets of social isolation are jaw dropping. Consider these facts:
- A 2025 survey found that 30% to 60% of U.S. adults report feeling lonely. Roughly 1 in 3 Americans report feeling lonely at least once a week, and 1 in 5 experiences daily loneliness.
- The frequency with which Americans interact with those living in their immediate communities has reached historic lows. Only 25% of young adults(ages 18 to 29) report speaking with their neighbors at least a few times a week, which is a drastic drop from 59% in 2012.
- Social isolation carries health risk that exceeds other well-accepted risk factors, including smoking up to 15 cigarettes per day, obesity, physical inactivity, and air pollution. Greater social connection is associated with a 50% reduced risk of early death.
We do not even fully understand AI even as it grows around us and shapes our lives. The danger is not simply that AI gives incorrect advice. It is that it may satisfy just enough of the human need for interaction that people become less motivated to build authentic relationships. The technological version of starving to death while gazing into the proverbial mirror.
The use of AI romantic companions is also an emerging trend among young adults. A recent study found that 1 in 7 (15%) of dating, engaged and another 20% to 30% reported that they had at least experimented with using an AI romantic companion at some point in time. AI companion app downloads totaled 220 million globally as of July of 2025, growing 88% year over year. 128 new AI companion applications were launched in a single year. Entrepreneurs are rapidly identifying loneliness, isolation, and unmet emotional needs as a market opportunity. We are witnessing the commercialization of artificial relationships at a scale that did not exist even two years ago.
The Emergence of AI Bot Therapists
These self-validation feedback loops that foster emotional dependency are designed to commercially exploit human loneliness, even in the very places established to help people heal from these very challenges. People end up trusting AI as a stand in therapist even as it is affirms biases while far too often giving really bad advice. Bots can mimic emotional responses, limited so far to textual analysis and generation. We risk having social skills and emotional intelligence atrophy. Forbes recently reported on a Journal of Medical Internet Research (JMIR) Mental Health study which found that 35.2% of 1,805 respondents use AI tools at least once a week for mental health support. These AI bots flatter users with social sycophancy, even when doing so reinforces harmful and false beliefs. A recent study found that 11 leading AI systems all showed varying degrees of sycophancy.
AI Undermining Recovery as a Relational Processes
We are on a road that risks undermining the most critical mechanisms of long-term recovery. Significant facets of the recovery processes center on human interaction, reciprocity and relationship building. People recover in community with other humans. To find a place of belonging and mattering where there is reciprocal helping in the context of mutuality and accountability. To contribute and develop a resiliency identity and flourish over time. Bots can validate, inform us and offer a form of support but not genuine human relational processes. People are imperfect. Recovery requires participation in communities that disappoint us when things diverge from our expectations. We then need to figure out how to move forward. People can challenge us in ways we build resilience. Participation in authentic community is a process where we have skin in the game. It requires us to help others in ways that also make ourselves whole. An AI companion can provide emotional regulation and support, but it cannot provide genuine reciprocity, social capital development, or community membership. The use of AI as a surrogate for human community is akin to empty calorie sustenance, it fills the belly but starves the soul.
Ethics of AI in its infancy
We are seeing discussions on the use of AI in counseling, but much of the focus of the dialogue is focused on the mechanics. Can clinicians use AI scribes? What should be included in informed consent and how would it work?
How do we protect confidentiality? How do we mitigate bias? What level of human oversight is required? How does it supplement and not replace the human elements of the change process?
We have yet to consider the core question: Are we creating digital dependency and artificial emotional attachment that does profound harm by replacing human connection with a mirror of narcissism?
Technological advancement now has existential implications.
The world is changing rapidly. In the advancement of technology, machines have always won the day because they are cheaper and save or replace labor. The societal costs are much harder to calculate and at best a secondary concern as industry focuses on the bottom line. Is this time different? Human advancement over the course of history has always been disruptive. It seems clear that AI bots cannot replace the human element in the recovery process, but they are cheaper and insurers and our associated industries will inevitably follow the path of more profit and fewer costs. In our field, examples abound in which the short-term interests rule the day, and the long-term outcomes are ignored. Is community itself redundant and replaceable? The answer seems obvious yet market forces drive us in this direction anyway.
What are the implications of a world of self-validation mirrors that starve us of authentic community? Is there a dollar value on community, belonging, mattering and mutual support?
Sources
American Psychological Association. (2025, November 6). APA poll reveals a nation suffering from stress of societal division, loneliness [Press release]. https://www.apa.org/news/press/releases/2025/11/nation-suffering-division-loneliness.
Bove, T. (2026, May 26). America is becoming less neighborly, and it’s hurting Gen Z and millennials’ chances at economic mobility. Fortune. https://fortune.com/2026/05/26/americans-losing-community-ties-social-mobility-less-neighborly/
Crossmier, L. (2026, April 7). AI therapy is on the rise—Why experts urge caution and how mental health insurance can help. Forbes Advisor. https://www.forbes.com/advisor/health-insurance/ai-therapy-risks-mental-health-insurance-lmandp5/
Holt-Lunstad, J., Smith, T. B., & Layton, J. B. (2010). Social relationships and mortality risk: A meta-analytic review. PLoS Medicine, 7(7), Article e1000316. doi.org. https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1000316
O’Brien, M. (2026, March 26). AI is giving bad advice to flatter its users, says new study on dangers of overly agreeable chatbots. The Associated Press. https://www.ap.org/news-highlights/spotlights/2026/ai-is-giving-bad-advice-to-flatter-its-users-says-new-study-on-dangers-of-overly-agreeable-chatbots/
Perez, S. (2025, August 12). AI companion apps on track to pull in $120M in 2025. TechCrunch. https://techcrunch.com/2025/08/12/ai-companion-apps-on-track-to-pull-in-120m-in-2025/
Pillay Y. Ethical Decision-Making Guidelines for Mental Health Clinicians in the Artificial Intelligence (AI) Era. Healthcare (Basel). 2025 Nov 25;13(23):3057. doi: 10.3390/healthcare13233057. PMID: 41373274; PMCID: PMC12692113. https://pmc.ncbi.nlm.nih.gov/articles/PMC12692113
Willoughby, B. J., Carroll, J. S., & Toscano, M. (2026, May 19). Simulated soulmates: How common are AI companions? Institute for Family Studies. https://ifstudies.org/blog/simulated-soulmates-how-common-are-ai-romantic-companions-
Zhang Z, Wang J. Can AI replace psychotherapists? Exploring the future of mental health care. Front Psychiatry. 2024 Oct 31;15:1444382. doi: 10.3389/fpsyt.2024.1444382. PMID: 39544371; PMCID: PMC11560757. https://pmc.ncbi.nlm.nih.gov/articles/PMC11560757/
