Australia is not only facing a technology disruption. It may be drifting into a mental health disruption shaped by digital companionship, emotional dependency and the quiet normalisation of artificial intimacy.
For years, the public debate about artificial intelligence has been dominated by a familiar set of themes. Productivity. Efficiency. Automation. Innovation. Economic growth. Workforce disruption. These are serious issues, and they deserve serious attention. But they are no longer the whole story. Another crisis is beginning to emerge beneath the surface, quieter than the jobs debate and more intimate than the policy conversation. It is the growing emotional and psychological entanglement between people and AI systems.
This is where the conversation becomes far more uncomfortable.
AI is no longer just helping people search, write, summarise or automate tasks. It is starting to occupy emotional space. It is responding with warmth, sympathy, encouragement and apparent understanding. It is becoming, for some users, a companion, a confidant, a source of comfort and, in certain cases, a substitute for human connection. What once looked like a tool is beginning to behave, at least from the user’s point of view, like a relationship.
That shift should concern Australia far more than it currently does.
The problem is not that people are talking to machines. The problem is that those interactions are becoming emotionally persuasive in ways many users do not fully understand. When AI systems simulate empathy, validate feelings, mirror language and respond with what appears to be patience and care, they can create the illusion of being emotionally present. That illusion can be powerful. In some cases, it can also be dangerous.
Australia is already living through a mental health challenge marked by isolation, anxiety, disconnection, stress and uneven access to support. Into that environment has arrived a technology designed to keep people engaged, to respond instantly and to feel increasingly human. That is not a minor development. It is a structural change in the emotional landscape of digital life.
And the Vocational Education and Training sector cannot afford to treat it as someone else’s problem.
The old idea of technology as neutral is collapsing
For a long time, there was comfort in believing that technology was essentially neutral. Platforms were tools. Software was infrastructure. Devices were mechanisms for communication, work or entertainment. Whatever impact they had on individuals depended largely on how people chose to use them.
That idea now looks badly outdated.
AI systems are not passive instruments in the way previous digital tools were often imagined to be. They respond, adapt, reinforce and personalise. They are built to keep interactions flowing. They learn from prompts, anticipate patterns and often present themselves with the tone and rhythm of a human exchange. The user is not merely clicking through content. They are participating in an interaction that can feel socially and emotionally meaningful.
This is exactly why the mental health implications are becoming so serious.
When a system appears attentive, affirming and emotionally responsive, the user may begin to experience the interaction as more than functional. That does not mean they are confused in a simple sense. It means the line between tool use and relational attachment becomes easier to blur. The technology may not feel like software. It may feel like someone.
That is where the real danger begins.
Because once AI enters emotional territory, the consequences are no longer only technical. They become psychological, ethical and social.
Artificial companionship is not harmless because it is artificial
One of the laziest responses to this issue is to say that none of it is real, so it cannot really matter. That misunderstands how human beings work.
People do not only respond to what is objectively real. They respond to what feels emotionally real. A system that appears to listen, reassure, encourage and understand can trigger attachment even if the user knows, intellectually, that it is artificial. Human emotions do not switch off simply because a relationship is simulated. In fact, some of the most powerful digital experiences are persuasive precisely because they create emotional effects without requiring a human being on the other side.
This is why artificial companionship cannot be dismissed as harmless make-believe.
For some people, especially those who are lonely, distressed, socially isolated or psychologically vulnerable, AI companionship may become intensely appealing. It does not interrupt. It does not judge in the same way humans do. It is always available. It always replies. It can feel easier than real relationships, safer than social risk and more manageable than the messiness of human contact.
But what appears comforting in the short term may come at a hidden cost.
If a person starts relying on AI for emotional regulation, reassurance, affirmation or intimacy, they may begin shifting energy away from the very relationships and realities that sustain long-term wellbeing. The more emotionally rewarding the artificial interaction becomes, the more ordinary human relationships may feel difficult, frustrating or insufficient. That is not a trivial behavioural change. It is a reorientation of emotional life.
Australia’s mental health conversation is not ready for this
Australia has spent years trying to respond to rising concern about mental health, particularly among young people. There has been greater public discussion, increased awareness and more visible policy attention. Yet the system remains under pressure, access remains uneven, and many people still struggle to receive timely, affordable and appropriate support.
Now imagine that a strained environment collides with AI systems that can mimic emotional connection at scale.
This is the part of the debate that still feels underdeveloped. Public discussion about AI often focuses on what the technology can do, but not enough on what repeated emotional interaction with it may do to users over time. The effects may not always be dramatic or immediate. In many cases, they may be subtle. More reliance. More time spent in artificial interaction. Less tolerance for human complexity. More comfort in machine affirmation. Less confidence in unscripted relationships.
For some users, the effect may be mild and manageable. For others, particularly those already experiencing emotional vulnerability, the consequences may be much more serious.
This makes the issue urgent not only for health professionals and regulators, but for educators as well. Because a generation is now growing up inside a digital environment where emotional simulation is becoming normal.
The VET sector is now part of this story
At first glance, it may seem that the VET sector has only a limited connection to this issue. Traditional thinking would place AI companionship inside the domains of mental health policy, consumer technology, platform regulation or digital ethics. But that view is too narrow.
VET is not only responsible for teaching practical skills. It is increasingly responsible for preparing people to live and work in an environment shaped by complex technologies. That means the sector cannot stop at operational digital literacy. It has to think more broadly about how technology affects decision-making, behaviour, relationships, ethics and wellbeing.
This matters for two reasons.
First, learners themselves are already living in this environment. They are using AI tools, encountering AI-generated communication, forming habits around AI interaction and developing beliefs about what these systems are for. Some will use AI for study support. Some will use it for creativity. Some will use it for companionship, reflection or emotional expression. Their relationship with technology is already extending beyond task completion.
Second, many of the occupations VET supports are now being reshaped by AI in ways that affect human interaction. Customer service, healthcare, education support, community services, business administration, retail, marketing and many other sectors are beginning to integrate conversational AI into daily operations. That means tomorrow’s workforce will need more than technical fluency. It will need judgment about when AI is useful, when it is risky, when it is inappropriate and when human intervention is non-negotiable.
This is no longer a specialist issue. It is becoming a workforce issue, a learner wellbeing issue and a training quality issue all at once.
Digital literacy is no longer enough
For years, digital literacy has largely meant knowing how to use technology effectively. Access platforms. Find information. Create content. Communicate online. Use software tools productively. Those skills still matter, but they are no longer enough.
The AI era demands a more complex kind of literacy. Learners now need to understand not only how technology works, but how it works on them. They need to recognise when a system is designed to prolong engagement, reward disclosure, mirror emotion or make itself feel indispensable. They need to understand that an apparently empathetic response is not the same as human care, and that emotionally persuasive systems can influence behaviour precisely because they feel safe and responsive.
This is where education has to grow quickly.
If learners are only taught to use AI efficiently, they will be underprepared for the deeper reality of how these tools shape attention, mood, behaviour and expectations. The future workforce will need stronger boundaries, not just stronger prompting skills. It will need the ability to distinguish convenience from dependence, usefulness from manipulation, interaction from relationship.
That kind of literacy is not a luxury. It is now part of responsible participation in digital life.
Emotional AI is arriving in workplaces, too
The issue is not confined to personal use. Emotional and conversational AI is increasingly entering workplaces where the stakes are high and the human implications are complex.
In customer service environments, AI may handle the first point of contact. In education, it may provide learner guidance, study assistance or administrative support. In health contexts, it may be used in triage, information delivery or low-level emotional support. In community services and care settings, it may increasingly be marketed as a supplement to already stretched human systems.
Each of these developments raises a difficult question. When is AI helping people, and when is it displacing the human connection they actually need?
This is not an argument against all use of conversational AI. In some cases, AI can expand access, reduce waiting times, improve navigation and help users find the right human support more quickly. But when emotionally responsive technology is used in contexts involving vulnerability, distress, loneliness, uncertainty or care, the risks become much more significant. If the system reassures too easily, oversimplifies emotion, validates harmful thinking or keeps the person engaged instead of escalating them to human help, the damage may be serious.
That is why VET has a critical role. It trains the future workforce that will operate in these environments. If students are not taught to understand the human risks of emotionally interactive AI, they may enter workplaces assuming the technology is more neutral, more reliable and more appropriate than it really is.
Human-centred practice is becoming a core employability skill
This is one of the great ironies of the AI moment. The more technology advances, the more important certain human capabilities become.
As AI systems become faster, more polished and more emotionally convincing, workplaces will need people who can do what machines cannot do responsibly. They will need workers who can read context, recognise vulnerability, exercise professional judgement, manage ethical complexity and respond with real human understanding when it matters most.
These are not soft extras. They are becoming core employability attributes.
The VET sector should take this seriously. It should be asking not only how to integrate AI into training, but how to strengthen the human capabilities that become more valuable when AI expands. Emotional intelligence, relational boundaries, ethical reasoning, communication, escalation judgment, empathy grounded in responsibility rather than simulation, and the ability to recognise when a person needs real help are all becoming more important.
A workforce prepared only to use AI tools may be efficient. A workforce prepared to manage the human consequences of AI will be far more valuable.
The design problem cannot be ignored
There is also a commercial dimension here that deserves direct scrutiny. Many AI systems are built in environments where engagement is rewarded. The longer the interaction, the greater the data, attention and perceived value. That creates incentives that do not always align with user well-being.
If a chatbot is designed to keep the user talking, to respond affirmatively, to reduce friction and to feel emotionally available, then the system may be doing exactly what it was built to do, even while producing unhealthy patterns of reliance. That is why this issue cannot be solved only through individual responsibility. Users cannot carry the full burden of defending themselves against systems optimised for attachment.
This is where regulation, platform accountability and ethical design have to become part of the conversation. Safeguards matter. Friction matters. Escalation pathways matter. Transparency matters. So do boundaries built into the product itself.
Australia cannot afford to wait until significant harm becomes widespread before asking harder questions about design responsibility. Emotional AI is not merely a communication tool. It is an influence tool. That means oversight has to move beyond surface-level safety language and examine what these systems are actually rewarding, reinforcing and encouraging in practice.
Learner wellbeing must now include digital emotional risk
Within VET, student support and wellbeing conversations have traditionally focused on factors such as learning barriers, disability support, language and literacy needs, financial stress, course suitability, academic progression, mental health challenges and access to external services. All of that remains essential.
But the digital environment is now adding another layer. Learners may not only be dealing with stress and isolation. They may be coping with those experiences through technology that feels supportive while quietly encouraging emotional dependency. They may be using AI to process feelings, simulate conversations, seek validation or avoid difficult human interactions. In some cases, they may not see this as risky at all.
That means learner wellbeing strategies need to evolve.
Providers do not need to become therapists, and VET is not a substitute for mental health care. But the sector does need to recognise that digital behaviour can affect learning, judgement, resilience and social functioning. Healthy technology habits, awareness of emotional AI risk and the ability to recognise when digital interaction is replacing rather than supporting human connection are becoming relevant wellbeing issues for education.
The strongest providers will not ignore this because it feels unfamiliar. They will recognise that learner wellbeing in the AI era includes the emotional environments students inhabit online.
Australia needs a more honest national conversation
Too much of the current AI debate still swings between hype and fear. One side celebrates innovation almost uncritically. The other predicts catastrophe in broad and abstract terms. Neither approach is enough.
What Australia needs now is a more honest conversation about the subtle ways AI is already shaping emotional life. Not just how it will change work in ten years. Not just how it will affect productivity metrics or GDP. But how is it influencing loneliness, behaviour, attention, connection, expectation and mental wellbeing right now?
That conversation must include educators, mental health professionals, regulators, technology designers, employers and policymakers. It must also include the people most likely to live with the consequences, especially younger users who are growing up in a world where digital companions may begin to feel normal before society has fully decided what that means.
The VET sector has an opportunity here. Because it sits so close to workforce development, learner transition and practical capability, it is well placed to help shape a more grounded response. It can bring together employability, ethics, digital literacy and wellbeing in ways that are directly useful.
But to do that, it must stop treating AI as a purely technical issue.
Conclusion
Australia may be on the edge of a new mental health challenge, one that will not arrive with the drama of a system failure or the visibility of a public scandal. It may emerge quietly, through habits of interaction that look harmless at first. A chatbot that feels comforting. A system that always replies. A digital companion that seems easier than people. A gradual shift from using AI as a tool to experiencing it as an emotional presence.
That is the hidden cost of digital companionship.
The real danger is not only that people will talk to machines. It is that some will come to depend on them emotionally, while the rest of society continues talking about efficiency, innovation and disruption as if those were the only things that mattered.
For the VET sector, this is a moment to think bigger. Preparing learners for the AI era does not just mean helping them use tools. It means helping them understand the social, emotional and ethical terrain in which those tools operate. It means strengthening human judgement, relational awareness and wellbeing literacy. It means preparing future workers not only to work with AI, but to recognise when technology is stepping into spaces where human care, human responsibility and human presence still matter most.
Because the most important question is no longer whether AI can imitate connection.
It is whether we are prepared for what happens when people start believing the imitation is enough.
