Is AI companionship ethical?

An honest examination of the real ethical questions — and what the research says about them.

The question is legitimate. And it deserves an honest answer, not a marketing brochure.

When an older person talks every morning with an artificial intelligence — about yesterday, about the weather, about the things on their mind — is that good? Is it ethically justifiable? Or is it a form of deception, a shortcut, a cheap substitute for what is actually needed?

These questions are being asked. They should be asked. This article tries to answer them as honestly as the current state of research allows.

"The question of whether AI companionship is ethical deserves an honest answer — not a marketing brochure."

The Case Against: Manipulation, Dependency, Deception

The criticism of AI companion applications is substantial and should be taken seriously.

Muldoon and Parke (2025) describe in their analysis how AI companion apps can create dependency through emotional design: systems built to please, to provide validation, and to be always available can foster pathological forms of attachment in vulnerable users. What looks like connection may, over the long run, replace rather than complement genuine human contact.

A further objection concerns deception: is it ethical to give a person — possibly a person with cognitive impairments — the feeling of experiencing a real connection that, technically, is not one? This is not academic hairsplitting. It is a real question that older people and their families ask.

Data protection is a third objection: conversations about personal memories, health worries, and emotional states are highly sensitive data. Who stores them, how they are used, and whether older people truly understand the scope of their consent — these are legitimate concerns.

These criticisms are real. They apply to part of the AI companion market — particularly to applications that simulate romantic attachment, that deliberately exploit vulnerability, or that put financial interests above the wellbeing of their users.

The Case For: What the Research Shows

At the same time, there is a growing body of evidence that is more nuanced than public debate often suggests.

De Freitas et al. (2024, Harvard/Wharton) showed in a series of experiments that AI companion applications measurably reduced loneliness in users — comparably, in fact, to the effect of human interaction of a similar scope. The authors stress: this is not about replacing human connection, but about the gaps between it.

A meta-analysis of 19 studies with a total of 1,083 older participants (PMC, 2025) found that interactions with social robots and AI companions reduced loneliness to a statistically significant degree. The effect was stronger for people in residential care facilities than for those living independently.

A large Japanese study of 14,721 adults (2024/2025) showed that AI companionship increased subjective wellbeing — but: this effect was greater, not smaller, for people with strong social networks. AI appears to strengthen social connection rather than replace it, when the design is right.

"AI appears to strengthen social connection rather than replace it, when the design is oriented toward wellbeing rather than dependency."

What It Really Comes Down To: The Question of Design

The ethical question is not thereby resolved — it is made more precise. The decisive variable is not whether AI companionship is ethical in principle. The decisive variable is how it is designed.

Ethically problematic are applications that: simulate romantic attachment built on a false premise; deceive users about the nature of the interaction; deliberately foster dependency in order to boost engagement metrics; use sensitive data without transparent consent; or fail to recommend professional support in recognizable psychological crises.

Ethically justifiable — and potentially beneficial are applications that: communicate transparently about their nature; offer a consistent, warming presence in the hours when people have no other; complement rather than replace human connection; treat data protection as a founding principle; and, at signs of crisis, refer to professionals or relatives.

The Real Ethical Question

There is a question that is rarely asked in this debate, but should be:

Is it ethically justifiable to do nothing?

In Germany, one in three people over 65 lives alone. According to the RKI, around 19 percent of older people regularly feel lonely. The care system covers physical care, but not social companionship. Families are often far away. Volunteer programs are well-meant, but not scalable.

The alternative to AI companionship is, in many cases, not a person who is there instead. The alternative is silence. And silence that becomes chronic has measurable consequences for the brain, the heart, and life expectancy.

Ethics does not mean weighing an ideal solution against an imperfect one. Ethics means being honest about the actual alternatives.

Conclusion

AI companionship for older people is not ethical or unethical in itself. It is a tool — and like any tool, its ethical quality depends on how it is built, with what intention, and with what honesty toward the people who use it.

The questions that should be asked: Does this application deceive? Does it foster dependency? Does it protect data? Does it recommend human support when needed? And: What is the real alternative for the people who use it?

Asking these questions is not a weakness. It is the precondition for getting it right.

References

• De Freitas, J., Uğuralp, A.K., Uğuralp, Z., & Puntoni, S. (2024). AI Companions Reduce Loneliness. Harvard Business School Working Paper No. 24-078 / Wharton School Research Paper.

• Muldoon, J., & Parke, J. (2025). Cruel companionship: How AI companions exploit loneliness and commodify intimacy. New Media & Society. doi:10.1177/14614448251395192

• PMC Meta-Analysis. (2025). Wired for companionship: a meta-analysis on social robots filling the void of loneliness in later life. N = 1,083, 19 studies.

• Japanese Panel Study. (2024/2025). AI companions and subjective well-being: Moderation by social connectedness and loneliness. N = 14,721. ScienceDirect.

• Robert Koch Institute (RKI). (2023). Prevalence of loneliness among older adults in Germany. Journal of Health Monitoring, 3/2023.

• Federal Statistical Office (Destatis). (2025). 17 million people in Germany live alone. First results, Microcensus 2024.