How is AI affecting the job market? What the research says in 2026
The short answer
Roughly one in four workers globally has a job affected by generative AI (ILO, 2025). But "affected" is not the same as "replaced." The research — from ILO, OECD, Anthropic, and others — points in the same direction: most jobs are changing, not disappearing. What's actually happening is that certain tasks are being automated while new ones emerge, shifting which competencies are in demand. If you're searching for a job right now, it's more relevant to understand how your role is changing than to worry about whether it still exists.
What do we mean by "AI exposure"?
There's a lot of talk about AI "threatening" jobs, but the research uses a more nuanced concept: exposure. It measures what proportion of an occupation's tasks AI could potentially affect — not how many jobs disappear.
Anthropic, the company behind the AI model Claude, published a study in April 2025 combining three data sources: occupation descriptions (O*NET), actual AI usage data, and theoretical capability assessments. They call the metric "observed exposure" — and it differs from earlier research by measuring what AI actually gets used for, not just what it theoretically could do (Anthropic, 2025).
The result: there's a large gap between what AI can do and what it actually does. Technical potential and real-world usage are not the same thing.
Which occupations are most exposed?
Anthropic's analysis shows that the most exposed occupations include programmers (75% task coverage), customer service agents, and data entry clerks (67%). ILO highlights media and web-related roles in particular, where generative AI's ability to create text, images, and video has significantly increased automation potential (ILO, 2025).
At the same time, roughly 30% of workers — in occupations like chef, mechanic, and bartender — have zero measurable exposure. Physical work, trades, and jobs requiring presence in a specific environment are essentially unaffected.
One pattern stands out: the most exposed workers earn on average 47% more, have higher education (17% with postgraduate degrees compared to 5% among non-exposed), and more often work in office and knowledge-intensive environments (Anthropic, 2025). In other words, it's not low-wage jobs most affected by generative AI — it's white-collar professions.
Changing, not disappearing
Perhaps the most important finding across all current research: there are no clear signs of increased unemployment among exposed occupational groups since ChatGPT launched in late 2022 (Anthropic, 2025). The OECD Employment Outlook confirms the same picture — despite rapid AI development, labour market demand for exposed occupations hasn't systematically declined (OECD, 2023).
ILO's 2025 update concludes that "most jobs will be transformed rather than made redundant" due to the continuing need for human judgement, creativity, and social interaction (ILO, 2025).
What is happening is that the content of roles is shifting. OECD's research shows that 72% of job postings in highly AI-exposed occupations require leadership, and 67% require knowledge of business processes (OECD, 2024). Social and strategic competencies are increasing in demand — routine administrative tasks are decreasing.
A warning signal: young people entering the workforce
A concerning signal in Anthropic's data: the hiring rate for young workers (22–25 years) in AI-exposed occupations appears to have dropped by roughly 14% after ChatGPT's launch (Anthropic, 2025). The researchers emphasise that alternative explanations exist — but the pattern is worth tracking.
If you're early in your career and targeting a highly exposed role (programming, marketing, administrative work), it's particularly valuable to complement with competencies AI can't easily replace: project management, client relationships, industry knowledge, and judgement in complex situations.
What it means for you as a job seeker
If you're searching for a job right now, the practical implication is straightforward: AI is changing what is in demand, not whether it is. That means:
Show that you can use AI, not that you're competing with it. Employers are looking for people who can use AI as a tool in their work — not people who do exactly what AI already does. If you're applying for a marketing role, mention how you use AI for research or content editing. If you're applying for a project management role, highlight your ability to lead teams through change.
Invest in competencies AI doesn't do well. Leadership, communication, negotiation, judgement in ambiguous situations, industry-specific knowledge. These are consistently the competencies increasing in demand in highly exposed occupations (OECD, 2024).
Follow your industry, not the headlines. "AI replaces 40% of all jobs" is a headline. The reality is that impact varies enormously between occupations, industries, and countries. Look at what's actually being asked for in the postings you're saving — that tells you more about your job market than any general study.
Sources
- Anthropic. (2025). "Labor Market Impacts of AI." anthropic.com/research/labor-market-impacts
- ILO. (2025). "Generative AI and Jobs: A 2025 Update." ilo.org/publications/generative-ai-and-jobs-2025-update
- OECD. (2024). "Who Will Be the Workers Most Affected by AI?" oecd.org
- OECD. (2023). "OECD Employment Outlook 2023: AI and the Labour Market." oecd.org