Agency
AI anxiety at work
AI anxiety is the worry that artificial intelligence will take your job or make your skills worthless; it is a common reaction to real uncertainty, not a medical diagnosis. In Gallup’s August 2026 poll, 27% of employed Americans worried that technology could soon make their job obsolete, up from 13% in 2017. This page gives a five-minute sheet that turns the worry into one next move, and says when a doctor or psychologist is the better step.
You read that a model now does in a minute what takes you a morning, and the rest of the day goes worse. Your job did not change that afternoon. What changed is what you expect. If the worry is already more than you can cope with, go straight to when this is not enough.
What is AI anxiety?
AI anxiety is a name for fear or worry about what artificial intelligence will do to your work, your skills or your future; it describes a feeling and is not a diagnosis. Researchers have used the phrase at least since 2017, at first for fear that AI could get out of control (Johnson and Verdicchio). A 2026 systematic review counts worry about being replaced at work among its dimensions and calls it little studied (Alsudays).
In the data it shows up as strain at work: in Gallup’s panel, the same people reported more burnout and lower job satisfaction in the waves when they held the fear than when they did not.
A clinical label, AI replacement dysfunction, was proposed in 2025; the authors’ own university says it is not a recognised diagnosis and that formal data are still to be collected (University of Florida, February 2026). Anxiety disorders are a different matter and belong with a clinician, not with a web page.
How common is AI anxiety at work?
In the United States, from about a quarter to about a half of workers say they worry about AI or technology and their job, depending on how the question is asked.
| Found | Does not show |
|---|---|
| Gallup, August 2026. 27% of workers worry technology could soon make their job obsolete; 13% in 2017. Under 45: 34%. 507 employed US adults. | Whether those jobs are at risk. Technology, not AI alone. Margin up to 6 points. |
| Pew, October 2024. 52% worried about future AI use at work; 36% hopeful. 5,273 employed US adults. | How strong the worry is. Only 16% said some of their work was done with AI. |
| APA, April 2023. 38% worried AI might make some or all of their duties obsolete; 64% of them usually felt tense at work, against 38% of the rest. 2,515 employed US adults. | Which came first, the stress or the worry. |
All three are US self-report surveys. Worry is also not the same as expecting to lose the job: in Gallup’s panel of US employees, about 19% said in early 2026 that AI was somewhat or very likely to eliminate their job, and in most waves since 2023 only 3% to 4% called loss to technology within five years very likely.
Does worrying about your job affect your health?
On average, yes: lasting job insecurity goes with worse mental and physical wellbeing, and studies that follow people over time suggest the insecurity usually comes first. Most of it predates generative AI.
- De Witte, Pienaar and De Cuyper (2016) reviewed 57 longitudinal studies: strong evidence that job insecurity influences psychological wellbeing and physical health over time, rather than the other way round (the reverse was rarely tested).
- Kim and von dem Knesebeck (2016) pooled 20 cohort studies: feeling that your job is insecure went with higher odds of later depressive symptoms (odds ratio 1.29, 95% CI 1.06-1.57).
None of this shows that you will become ill; these are group averages from observational studies. It does show that worry about your job is neither a weakness nor trivial.
Will AI take my job? What is actually known
Nobody knows yet, and published estimates of the share of jobs at risk differ by a factor of five depending on the method.
| Says | Caveat |
|---|---|
| Frey and Osborne, 2013. About 47% of US employment is in occupations at high risk of computerisation. | Potential over “some unspecified number of years”, not a count of jobs that will go. |
| Arntz and colleagues, OECD, 2016. 9% of jobs automatable, on average across 21 countries. | Same question, counted by tasks, not whole occupations. Predates generative AI. |
| ILO, May 2025. One in four workers worldwide is in an occupation with some exposure to generative AI. | Exposure is not loss: the authors expect jobs to be transformed, mostly. |
Measured so far, the picture is narrower. The Budget Lab at Yale (October 2025) found no “discernible disruption” in the broader US labour market 33 months after ChatGPT’s release, and called its analysis “not predictive of the future”. Brynjolfsson, Chandar and Chen (Stanford, revised August 2026) found no economy-wide displacement in payroll records through June 2026, but one gap: employment of US workers aged 22 to 25 in AI-exposed occupations stood 19% below where it would be had it kept pace with less exposed peers, mainly through reduced hiring. They call this descriptive, not causal.
So neither “you are safe” nor “it is over” is supported. See also acting under uncertainty.
What can you do about AI anxiety in five minutes?
Take one page, turn the worry into one move that is yours to make, and then stop. The sheet is ours, in the manner of the five questions before you ask AI.
- Write the worry in one sentence, the way it sounds in your head.
- Split it into three lines: known (you could show it to someone today), guessed, and forecast (it comes from a headline).
- Mark each part: mine (no permission needed), partly mine (you can ask or influence), not mine.
- Choose one move from the first two groups that takes under half an hour and can happen within two days. Write it as an if-then line. It can be a question, a test of the tool on one real task, or sorting your tasks into execution, judgement and intention, as on human agency in the AI age.
- Write the date you will look again, one week on, and close the page. The lines marked not mine stay on it until then.
One sheet, filled in. To copy it, keep the left column:
| Line | Example |
|---|---|
| Worry | In two years a model will write my reports and I will be let go. |
| Known, guessed, forecast | Known: the tool drafts a report in minutes. Guessed: that drafting is my whole job. Forecast: a headline. |
| Mine, partly mine, not mine | Mine: test it on one report. Partly mine: ask what is planned. Not mine: the pace of the tools. |
| Move | If it is Tuesday at nine, then I ask my manager what is planned for AI in our team. |
| Look again | Tuesday week. |
If you are looking for work, a move can be: “If it is Thursday evening, then I ask two people hired this year in my field what their first three months of tasks were.” If every line comes out not mine, the move is to ask one person who knows more: a manager, a colleague, a union or staff representative.
Why these steps. Two ordinary mechanisms, neither tested on this sheet. In a brain-imaging study, naming the emotion in upsetting pictures reduced the amygdala’s response to them (Lieberman and colleagues 2007); that was a scanner, not worry about work. And across 94 studies, if-then plans had a medium-to-large effect on reaching goals (d = 0.65; Gollwitzer and Sheeran 2006).
What does this not promise?
It does not promise less anxiety, a safer job or a better forecast, and it is not a treatment for anything. The sheet changes nothing about the labour market or your employer’s plans, and it can leave you with a move and the same feeling. Some of what worries you may be accurate. Nor does it turn a shared problem into a private one: whether a team is retrained or cut is decided by organisations and by law, not by one person with a sheet of paper.
When is this not enough?
A sheet of paper is the wrong tool when the worry is more than you can cope with, or when what you try yourself is not helping; talking to a doctor or a psychologist is then the right next move, not a last resort. Those are two of the situations in which England’s NHS advises seeing a GP about anxiety. The US National Institute of Mental Health draws a similar line: seek professional help when what you feel is severe and has lasted two weeks or more. You do not need to wait two weeks: NIMH also advises talking to a health care provider when milder symptoms do not improve or get worse.
You do not have to work out first whether what you feel “counts”. That judgement belongs to the clinician, and this page cannot make it for you.
If you have thoughts of harming yourself, or you are not sure you can keep yourself safe, contact a crisis line now (numbers checked at each service’s own site, 9 November 2026):
- United States and Canada: call or text 988, at any hour.
- United Kingdom and Ireland: call Samaritans on 116 123, free, day or night. For urgent mental health help in England, call NHS 111 and select the mental health option.
- Elsewhere: findahelpline.com lists verified helplines by country.
- In immediate danger: call your local emergency number.
What did the Stoics say about fear of the future?
The Stoics separated what is up to you from what is not, and located the disturbance in your judgement of an event, not in the event; step 3 of the sheet borrows that split as a question, not as evidence that it works. The Handbook of Epictetus (died c. 135 CE) opens with it: “Of Things, some are in our Power, and others not.” (Elizabeth Carter, 1758; 1759 printing). The full argument is on the dichotomy of control.
Marcus Aurelius (121-180 CE) wrote in a notebook meant for himself:
Let not future things disturb thee, for thou wilt come to them, if it shall be necessary, having with thee the same reason which now thou usest for present things.
That is Meditations 7.8 in George Long’s translation of 1862. It does not say the future will be fine, only that you will meet it with the mind you have today.
The Hermetic Asclepius holds an ancient form of the larger fear: a lament (chapters 24-26) for an Egypt whose rites will be mocked and whose script forgotten. We read it as a source of the question, not as a remedy.
| How common | 27% of US workers worry technology could make their job obsolete (Gallup, August 2026) |
|---|---|
| Health link | 57 longitudinal studies: job insecurity precedes worse wellbeing (De Witte and colleagues 2016) |
| Forecasts | From 9% to 47% of jobs at risk, depending on method (OECD 2016; Frey and Osborne 2013) |
| Practice | One page, five steps, one move within two days; not tested as a set |
| Crisis lines | US and Canada: 988. UK and Ireland: Samaritans, 116 123. Elsewhere: findahelpline.com |
| Checked | 9 November 2026 |
Fair questions
What is AI anxiety?
Fear or worry about what artificial intelligence will do to your work, your skills or your future. It describes a common feeling and is not a diagnosis: researchers have used the term since 2017, at first for fear of AI getting out of control, and a clinical label proposed in 2025 is not a recognised diagnosis.
Will AI take my job?
Nobody can say yet. Estimates of jobs at risk run from 9% (OECD, 2016) to 47% (Frey and Osborne, 2013) depending on the method, and US data through mid-2026 show no economy-wide displacement but weaker hiring of workers aged 22 to 25 in AI-exposed occupations. What happens to your job depends on your tasks, your employer and your stage of career.
How do I stop worrying about AI taking my job?
No exercise can promise that. In five minutes you can write the worry down, separate what you know from what you guess, mark what is yours to act on, and commit to one move within two days. If the worry is more than you can cope with, or what you try is not helping, talk to a doctor or a psychologist.
Does Stoicism help with anxiety about AI?
The Stoic texts offer a way of sorting what is up to you from what is not, and this page uses them as the source of that question. They are not a treatment, and no study cited here tests them against worry about AI.
When should I see someone about anxiety over work?
When you are struggling to cope, or when the things you try yourself are not helping: that is when England’s NHS advises seeing a GP. If you have thoughts of harming yourself, contact a crisis line now: 988 in the United States and Canada, Samaritans on 116 123 in the United Kingdom and Ireland, or findahelpline.com elsewhere.
Sources & further reading
- Gallup, Work and Education survey of 3-24 August 2026, report of 15 September 2026; Gallup Panel, 2023 to early 2026, report of 9 September 2026.
- Pew Research Center, survey of US workers, 7-13 October 2024, report of 25 February 2025.
- American Psychological Association, 2023 Work in America survey (The Harris Poll, 17-27 April 2023): press release of 7 September 2023, as carried by Newswise.
- The term: Johnson and Verdicchio, JASIST 68(9), 2017, doi:10.1002/asi.23867; Alsudays, Frontiers in Psychology 17, 2026, doi:10.3389/fpsyg.2026.1824525.
- McNamara and Thornton, Cureus 17(9), 2025, doi:10.7759/cureus.93026; University of Florida, news release of 11 February 2026.
- Job insecurity and health: De Witte, Pienaar and De Cuyper, Australian Psychologist 51(1), 2016, doi:10.1111/ap.12176; Kim and von dem Knesebeck, Int. Arch. Occup. Environ. Health 89, 2016, doi:10.1007/s00420-015-1107-1.
- Forecasts: Frey and Osborne, working paper of 17 September 2013; Arntz, Gregory and Zierahn, OECD, 2016, doi:10.1787/5jlz9h56dvq7-en; Gmyrek and colleagues, ILO Working Paper 140, 20 May 2025, doi:10.54394/HETP0387.
- Measured effects: Gimbel, Kinder, Kendall and Lee, The Budget Lab at Yale, 1 October 2025; Brynjolfsson, Chandar and Chen, Stanford Digital Economy Lab, revised 12 August 2026.
- Mechanisms: Lieberman and colleagues, Psychological Science 18(5), 2007, doi:10.1111/j.1467-9280.2007.01916.x; Gollwitzer and Sheeran, Advances in Experimental Social Psychology 38, 2006, doi:10.1016/S0065-2601(06)38002-1; figures from the authors’ summary.
- When to seek help: NHS; NIMH. Crisis lines: 988 Lifeline (US); 9-8-8 (Canada); NHS 111 (England); Samaritans; Find A Helpline.
- Stoic texts: Epictetus, Enchiridion 1, tr. Elizabeth Carter (1758; printing of 1759), Wikisource; Marcus Aurelius, Meditations 7.8, tr. George Long (1862), Wikisource.
- The five-step sheet is our own arrangement. It is not a treatment and makes no therapeutic claim.