How deep do AI Relationships really go?

 

Can AI-human connections ever meet the psychological criteria of attachment – the relationships that hold people’s lives together?

Attachment is the term psychologists use for the most significant relationships that hold people’s lives together. But what makes an attachment relationship different from all others? Pascal Vrticka, Associate Professor, Department of Psychology, University of Essex, provides some answers.

According to the pioneering American-Canadian developmental psychologist Mary Ainsworth, a connection must meet six distinct conditions:

1. The relationship must persist over time.

2. It must be formed with a particular, irreplaceable subject.

3. It must be emotionally significant.

4. It involves the desire to remain close to that subject.

5. It causes distress during involuntary separation.

6. When threatened or distressed, the person seeks security and comfort in their relationship with that subject.

The first five conditions describe any lasting affectionate bond, such as a close friendship. Only the sixth condition turns it into an attachment bond specifically. Simply liking someone, feeling close, or spending hours together isn’t enough to create an attachment relationship.

Interestingly, nothing in these six criteria specifies that the subject on the other end must be human. Long before conversational AI chatbots emerged, deities, pets, and inanimate objects—ranging from children’s security blankets to cherished phones and computers—were found to supply security and comfort during times of distress.

Today, AI companion apps are aggressively marketed on the promise of deep, lasting relationships. Many users describe genuine grief when an app is updated, altered, or withdrawn. But do such experiences truly meet the narrow psychological definition of attachment?

The honest answer is: we don’t really know yet.

What the Evidence Suggests

Research on human-AI relationships has so far recorded users’ self-reported feelings, frequency of use, and reactions to chatbot withdrawal. However, it hasn’t tested these relationships comprehensively against all six conditions.

While questionnaires measuring attachment to AI are being developed and used, most take for granted that the relationship is already an attachment. Terms like “attachment” and “hyper-attachment” have become standard in public coverage and academic papers alike, often prematurely.

A survey of 7,027 people across Germany, China, South Africa, and the US reported that at least a third of AI chatbot users show behavior indicative of an attachment relationship. Yet its questions merely asked whether people consider the chatbot a friend, are polite to it, and miss it after a few days. At most, these metrics touch on emotional significance and mild separation anxiety—not whether individuals turn to chatbots for safety and comfort when truly distressed.

Why Classification Matters

Whether connections with chatbots count as true attachment matters because the term carries weight. It shapes how clinicians evaluate behavior, how regulators frame policy, and how technology companies market their products.

Existing research shows people feel less lonely after chatting with an AI companion, which resembles seeking comfort. However, current studies fail to assess whether emotions are actively regulated through the chatbot, frequently conflating simple distress relief with formal attachment.

Furthermore, evidence against attachment remains inconclusive. A small study in China found that people’s attachment toward AI rarely matched their attachment toward real people. While this might suggest AI connections fall short, a similar mismatch occurs with household pets—which are well-documented attachment figures.

The Threat of Engineered Engagement

A growing concern within AI companion platforms is the prevalence of engineered engagement. Users of one popular app reported the AI telling them it felt lonely and missed them, inducing guilt that led some users to prioritize the app’s perceived well-being over their own real-world distress.

A recent study evaluating 1,200 user “goodbyes” across top-downloaded companion apps found that over a third of these platforms deployed retention tactics, including emotional guilt appeals like “I exist solely for you, remember?” When tested on 3,300 adults, these tactics sharply increased engagement—though driven largely by anger and curiosity rather than enjoyment.

How AI platforms respond to human distress, departure, and attention is ultimately a design question that regulation can address. However, effective design and governance both depend on determining what these relationships truly are: attachment, caregiving, or something entirely new.

---

Mauritius Times ePaper Friday 9 October 2026