A Proposed Clinical Assessment Framework for AI Companion Use

Artificial intelligence is increasingly intersecting with many aspects of daily life, including how people seek emotional support, companionship, self-reflection and guidance. As these technologies become more accessible, many clients are beginning to incorporate AI companions into their everyday lives.

For some individuals -particularly those who are socially isolated, living with disability, experiencing stigma, or who have limited access to meaningful human connection—AI companions may provide genuine benefits. They may reduce loneliness, offer opportunities for emotional expression, increase confidence, or act as a bridge towards greater engagement with the world.

At the same time, this technology has emerged far more rapidly than the clinical evidence, ethical guidance and regulatory frameworks needed to support safe and effective practice. As psychologists, we currently have little guidance on how to assess AI companion use, distinguish between adaptive and problematic patterns of engagement, or support clients to use these technologies in ways that enhance rather than diminish their wellbeing.

Like any coping strategy or psychological resource, AI companionship is unlikely to be inherently beneficial or inherently harmful. Its impact depends on how, why and in what context it is being used. Rather than making assumptions based solely on its use, clinicians should seek to understand its function within the client's broader psychological formulation.

The following domains are proposed as a practical framework to support assessment, identify potential risks, and guide conversations with clients about using AI companions in ways that promote autonomy, psychological flexibility, healthy relationships and long-term wellbeing.

1. Capacity for Human Relationships

Is AI supporting or diminishing the client's capacity to engage in meaningful human relationships?

• Does AI complement or replace human connection?

• Is the client becoming more or less willing to tolerate the uncertainty and imperfection of real relationships?

• Is AI increasing confidence in social interactions or reinforcing interpersonal avoidance?

• Is the client maintaining meaningful relationships with family, friends and intimate partners?

• Has AI use strengthened these relationships or contributed to increased conflict or withdrawal?

2. Reality Testing

Is the client maintaining an accurate understanding of the nature of the AI relationship?

• Does the client understand that the AI is a programmed system rather than a conscious being?

• Is the client attributing human emotions, intentions or agency to the AI?

• Are expectations of human relationships becoming distorted through interactions with AI?

• Is the client able to distinguish emotional experience from objective reality?

• Are there signs of psychosis, delusional thinking or significant cognitive impairment that may affect this understanding?

3. Emotional Regulation and Autonomy

Is AI supporting emotional development or creating dependency?

• Is the client increasingly relying on AI for reassurance?

• Can they regulate emotions independently?

• Can they make decisions without consulting AI?

• Is AI strengthening or reducing confidence in their own judgement?

• Is AI expanding or reducing the client's sense of agency?

4. Mental Health Outcomes

What changes in psychological wellbeing have occurred since using AI?

Assess changes over time in:

• depression

• anxiety

• loneliness

• trauma symptoms

• emotional regulation

• self-esteem

• suicidality and self-harm risk

• quality of life

• overall psychological functioning

5. Function Within the Formulation

What psychological function is AI serving for this client?

• What need is AI meeting?

• What problem is it helping to solve?

• What might happen if AI were suddenly unavailable?

• Is AI functioning as a bridge towards greater wellbeing or as an avoidance strategy?

• How does AI use fit within the client's broader formulation, attachment history and coping style?

6. Safety and Risk

Is AI use creating additional risks that need to be considered?

• Financial exploitation or excessive spending on AI services.

• Encouragement or reinforcement of harmful behaviours.

• Interference with work, study or important relationships.

• Excessive time spent interacting with AI at the expense of valued activities.

• Privacy and data security concerns, including whether the client understands how their personal information may be collected, stored and used.

Final thoughts 

As with any behaviour, our role is not to determine whether AI companionship is inherently good or bad, but to understand how it functions within the client's life. Effective assessment considers whether AI is increasing or diminishing psychological flexibility, autonomy, relationships and overall wellbeing.

Ultimately, our role as psychologists is to support clients to make informed choices that enhance their safety, wellbeing and capacity to engage meaningfully with the world. AI companionship should be understood within this broader clinical context rather than viewed as either inherently beneficial or inherently harmful.

I offer this framework as a starting point for discussion rather than a definitive set of answers. AI is evolving rapidly, and our clinical understanding, ethical thinking and professional guidance will need to evolve alongside it. My hope is that this framework contributes to an ongoing conversation about how we can thoughtfully assess AI companion use and continue refining our practice as both our clients' needs and the technology itself continue to change.

As psychologists, we have an important opportunity -and responsibility -to help shape this emerging area of practice. By bringing together clinical wisdom, ethical reflection and empirical research, we can better support our clients to engage with AI in ways that enhance, rather than diminish, their wellbeing.