Marcellyn Tadiwanashe Tagwireyi
As artificial intelligence becomes an increasingly accessible source of emotional support, the challenge is to use its convenience without mistaking simulated understanding for accountable psychological care.
When AI Became the First One I Told
The first time I struggled to understand my feelings after a relationship ended without closure, I turned to AI.
I described the pain and confusion caused by the way I had been treated, as well as the attachment I still felt.
I was reluctant to share these feelings with anyone for fear of judgment.
What made the interaction appealing was not only the speed of the response, but also the understanding I felt I received.
The chatbot responded immediately, focused on my current emotions, and validated my experience without blaming me for staying attached to an unhealthy relationship.
The experience showed me why people who feel isolated, ashamed or unable to afford therapy may turn to AI chatbots such as Gemini or ChatGPT.
AI does not become tired, appear shocked or ask users to wait for the next available appointment.
Yet the same qualities that make it convenient may encourage users to trust it beyond what its capabilities justify.
The danger begins when an accessible and supportive conversation is mistaken for a therapeutic relationship.
The Therapist Who Never Sleeps
The growing appeal of AI mental-health support is understandable.
For someone distressed, lonely or unable to access counselling, a chatbot can feel like a patient listener who is always available.
Researchers call this techno-emotional projection.
This is when users place human needs and expectations onto a responsive but non-conscious system.
Through repeated conversations, this may develop into a synthetic attachment, which is a connection that feels real although the machine neither feels nor understands.
Yet AI offers practical benefits. It can provide psychoeducation, coping strategies and important reflective questions within seconds, without appearing tired or judgemental.
This is especially valuable where therapy is constrained by cost, stigma or shortages of trained professionals.
However, availability is not clinical competence. An immediate answer may misread distress, overlook danger or confidently give inaccurate advice.
The question is not whether AI can sound therapeutic, but whether sounding therapeutic makes it a therapist.
When the Algorithm Says “I Understand”
When an AI chatbot says, “I understand,” its response may sound caring, but it does not arise from concern, experience or emotional awareness.
This is artificial empathy: the ability to recognise emotional cues in text, tone or facial expressions and generate an empathic response based on learned data.
Although technologically impressive, such empathy lacks the depth, intentionality and cultural sensitivity needed for genuine emotional resonance.
A chatbot cannot form a holistic understanding of how grief, family relationships, faith, poverty or past trauma shape a person’s present distress.
It may also misinterpret emotions expressed differently across cultures; for example, mistaking silence for disengagement when it may communicate respect, pain or reflection.
Most importantly, AI has no lived experience from which to connect with human vulnerability.
It can reproduce the language used for compassion without actually feeling concern for the person receiving it.
In mental-health care, this distinction matters because clients need more than reassuring words; they need to feel understood within the full context of their lives, culture and relationships.
Convenient, but Who Is Responsible When It Goes Wrong?
Using AI therapeutically raises questions about boundaries, transference and clinical responsibility, particularly when distress develops into a psychological crisis.
A person experiencing suicidal thoughts may communicate danger indirectly through statements such as “Why am I still here?” or even “Life seems to drag on.”
A psychologist can explore intent, planning, access to means, protective factors and immediate safety, while responding to tone, hesitation and context.
AI may offer emergency contacts or evidence-based coping techniques, such as the 5-4-3-2-1 grounding exercise, to redirect attention to the present environment during overwhelming moments.
Although this may reduce distress, it cannot replace risk assessment, physically secure the person’s safety or coordinate urgent intervention. Similar concerns arise with psychosis, intoxication, abuse or severe withdrawal.
Confidentiality presents another issue. Users may disclose experiences of trauma or sexuality without understanding how their information is stored, processed or shared.
Typing in private should not be confused with ethically protected confidentiality of professional therapy.
Psychologists are accountable to clients, professional ethics, regulatory bodies such as the Allied Health Practitioners Council of Zimbabwe, and professional associations like the Zimbabwe Psychological Association.
If AI provides harmful advice, responsibility may be divided among the developer, platform and user.
The system itself cannot accept moral responsibility, experience remorse or participate in professional disciplinary action, which is an alarming limitation when its user is vulnerable.
A Tool for Psychologists, Not a Replacement
Recognising these risks does not require psychologists to reject AI.
Historically, psychologists have been rightly wary of technology.
However, as APA Services observes, AI is already shaping psychologists’ work. When it used carefully, it can help organise information, identify patterns in data and expand access to psychological resources.
For overstretched practitioners, AI may reduce time spent on routine work, leaving more time for formulation, ethical reflection and engagement with clients.
However, assistance should not become substitution. Psychologists must critically evaluate AI-generated material rather than treating it as authoritative.
They should protect client information, understand how the tool uses data, consider possible cultural bias and obtain informed consent where AI becomes part of care.
Any recommendation must still be interpreted through the client’s history, relationships, culture and circumstances.
AI may help a psychologist prepare questions or summarise information, but the practitioner remains responsible for judgement and consequences.
The most responsible future, therefore, is psychologists using technology transparently while safeguarding human connection, professional accountability, and client safety.
Mental Health Care Still Needs a Human Face
Looking back at my own interaction with AI, I recognise both its usefulness and its limitations.
The experience was supportive, but it did not provide the accountability or relational depth of therapy.
AI will probably become increasingly present in psychological practice.
The challenge is not to prevent innovation, but to ensure convenience does not replace competence, confidentiality, and human connection.
People deserve mental-health support that is accessible, but they also deserve to know who is responsible for the care they receive.
AI may have a place in the consulting room, but it should never become the consulting room.
