When the Algorithm Feels Like a Verdict: Working Well in the Age of AI Anxiety

Written By Addy Dunkley-Smith

By Dr Addy Dunkley-Smith, Doctoral Clinical Psychologist | May 2026 | 4-minute read

Priya's 3am

Priya is 34, a senior marketing manager in Melbourne. She has built her career on the kind of careful, considered copywriting her clients hire her for. Last Tuesday her CEO emailed the team a screenshot of a generative AI tool producing a campaign brief in fourteen seconds. The message ended with three smiling faces. By Thursday Priya was waking at 3am with a tight chest, scrolling job ads on her phone. By Friday she could not bring herself to open her laptop. "I keep waiting to be told I am obsolete," she told her GP.

Priya is not unwell. She is responding, normally and humanly, to one of the most pervasive psychological shifts of the past year.

What the Research Is Calling This

Across recent global surveys, 69% of workers believe AI will lead to layoffs at their company within three years, and 49% fear they will personally lose their job to AI (ADP Research Institute, 2026; Fair Play Talks, 2026). Almost a quarter report that AI is already negatively affecting their mental health.

Researchers now use the term AI anxiety to describe the cluster of cognitive, emotional and behavioural responses linked to working alongside increasingly capable AI systems. Recent peer-reviewed work links it to:

  • Heightened technostress, mental exhaustion and information overload (Shalu et al., 2026, Annals of Neurosciences).

  • Loss of self-assurance in decision-making and reduced creative output (Mitigating AI anxiety on creativity, Springer, 2025).

  • Career uncertainty, emotional exhaustion and counterproductive work behaviour (Frontiers in Psychology, 2025).

  • Increased turnover intention partly mediated by "quiet quitting" (Erkmen & colleagues, PMC11939379, 2025).

A University of Florida group has gone further and proposed a clinical descriptor, AI Replacement Dysfunction (AIRD), characterised by anxiety, insomnia, loss of identity, feelings of worthlessness and hopelessness in response to perceived AI-driven job loss (University of Florida, 2026).

This is not paranoia, and it is not a personal failing. The threat is structurally ambiguous. Workers are asked to use the same tools they fear may replace them, often with no transparency about how their employer will respond. Quite frankly, in this rapidly changing arena, their employer may not even know themselves. Ambiguous, uncontrollable threat is precisely the kind of stressor that primes the nervous system for chronic worry.

Four Evidence-Based Steps to Try This Week

Most of the well-evidenced workplace interventions for distress under uncertainty draw on Acceptance and Commitment Therapy (ACT) and cognitive behavioural principles. A meta-analysis of randomised trials of ACT in the workplace found small to moderate reductions in psychological distress, anxiety and burnout, alongside improvements in mental health and work functioning (Unruh, Bricker et al., 2022, Journal of Contextual Behavioral Science). A four-session workplace ACT programme reduced distress and burnout in healthcare staff (Prudenzi et al., 2022, PLOS ONE), and ACT has reduced work-related rumination and job fatigue in operational staff (Mohammadi et al., 2024, BMC Psychiatry).

1. Name the thought; do not obey it. When your mind says "I will be obsolete by Christmas", that is a thought, not a forecast. Cognitive defusion exercises (silently labelling, "I am having the thought that..." before the worry) reduce the believability and behavioural pull of unhelpful cognitions. Practise it twice a day for a week and watch whether you act on the worry less often.

2. Reduce the inputs that keep the threat live. Repeated exposure to AI-layoff news amplifies catastrophic appraisals and sustains physiological arousal (Spagnoli et al., 2025, Frontiers in Psychology). If necessary, schedule a short daily window for AI news. Outside that window of time, the tab stays closed. Treat it as a one-week behavioural experiment, not a permanent rule.

3. Move toward values, not away from fear. ACT's strongest predictor of recovery is committed action in line with values (Vega-Campos et al., 2025, SAGE Open). Ask yourself: if AI anxiety were not steering me, what kind of professional would I want to be over the next twelve months? Translate the answer into one small weekly behaviour. A conversation with a mentor. An hour of upskilling. A creative project that uses your distinctly human strengths. Workers who feel their employer is investing in their skills are more than five times as likely to report job security (ADP Research Institute, 2026). The same logic applies when you invest in yourself. Find your values using our reflective tool HERE.

4. Ask for the conversation you are avoiding. Schema therapy work on the vulnerability to harm and failure early maladaptive schemas (i.e. belief systems formed through dealing with difficult life experiences that have become outdated and unhelpful) suggests that ambiguity feeds the schema; specificity disarms it. Where possible, ask your manager a concrete question: which parts of my role does the organisation expect AI to change in the next year, and what is the plan for the rest? Even an imperfect answer reduces the cognitive load of having to generate worst-case scenarios.

If you notice persistent insomnia, intrusive worry, loss of identity or hopelessness lasting more than a fortnight, that is the point at which to see a registered psychologist or your GP. AI anxiety is not yet a formal diagnostic category, but the conditions it commonly co-occurs with (generalised anxiety, adjustment disorder, depression) are treatable.

The Bigger Picture

The most useful reframe may be the simplest one. The discomfort you feel is information, not weakness. It is telling you that your work matters to you, that your livelihood matters to you, and that something genuinely uncertain is happening in your industry. The aim is not to feel nothing. The aim is to keep doing what matters while you feel it.

Priya rebooked her annual leave, capped her LinkedIn time at fifteen minutes a day, and signed up for a short course in audience research, a part of her work that AI cannot yet do well. She still has the 3am moments. She has them less often, and she gets back to sleep faster.

References and Further Reading

Addy Dunkley-Smith
Next

Silent Burnout Is the Workplace Crisis Nobody's Talking About — Until It's Too Late