← Back to full report
What This Report Is About
This report compares how ordinary people lived before and after ChatGPT changed the world, using roughly 2015 to 2022 as the pre-AI era and 2023 to 2026 as the post-AI era. It covers seven areas: life satisfaction, daily happiness, perceived freedom, stress, job security, who is most affected, and whether AI lets non-experts do work once reserved for specialists.
The aim is to separate measurable change from hype. Where the data cannot support a claim, the report says so, especially where changes could reflect the pandemic recovery or the cost-of-living shock rather than AI.
Key Findings
The productivity gains are real, and the biggest gains often go to the least experienced workers. In randomised experiments, ChatGPT made professional writing 40 percent faster (Noy and Zhang 2023), novice customer support agents resolved 34 percent more issues per hour (Brynjolfsson et al. 2025), and software developers completed 26 percent more tasks (Cui et al. 2026). These are the strongest measured effects in the whole report.
Global wellbeing has not visibly moved with the ChatGPT boundary. The world average life evaluation rose from 5.38 in 2015 to 5.65 in 2025, but most of the rise happened in 2020 and 2021, before ChatGPT. The Negative Experience Index fell from a record 33 in 2021 to 31 in 2024, and population stress fell from 41 to 37 percent, trends that fit the pandemic recovery better than an AI story.
The risks are concentrated where averages cannot see them. Only about 5.5 percent of employment in high-income countries sits in jobs highly exposed to AI automation (ILO), but women's employment is more than twice as exposed as men's, workplace stress stays high at 41 percent, and students with unstructured ChatGPT access did 17 percent worse on an exam without the tool (Bastani et al. 2024).
Key Statistics
ChatGPT grew from 100 million weekly users in November 2023 to about 800 million by October 2025
US adults who have used ChatGPT: 18% in 2023, 44% in 2026; 58% of 18 to 29 year olds by 2025 (Pew)
World average life evaluation: 5.38 in 2015, 5.65 in 2025 (Cantril ladder, World Happiness Report)
Positive Experience Index recovered from 69 in 2021 to 72 in 2024; Negative Index fell from 33 to 31
Daily stress among all adults: 41% in 2021, 37% in 2024; employee stress: 41% in 2024 (Gallup)
IMF: about 60% of jobs in advanced economies are exposed to AI; ILO: 5.5% highly exposed in high-income countries
WEF 2025 projects 170 million jobs created and 92 million displaced by 2030, a net gain of 78 million
88% of organisations used AI in at least one function in 2025, up from 55% in 2023 (McKinsey)
75% of knowledge workers used generative AI at work in 2024; 46% reported burnout (Microsoft)
Unstructured ChatGPT access: 48% better on practice problems but 17% worse on the no-AI exam (Bastani et al.)
The Non-Expert Question
Can AI let non-experts do expert work? The answer so far is: sometimes, and it depends on the task. On writing, coding, and customer support, the least experienced workers gained the most, which looks like genuine democratisation. On complex entrepreneurial decisions, the opposite happened: high performers gained about 15 percent while low performers lost about 10 percent, because they could not judge the AI's advice. And on learning, using AI as a crutch reduced exam performance by 17 percent.
The report treats this as an open question. The same tool can compress the cost of expertise and quietly remove the practice that builds it, and which effect wins depends on task structure, feedback, and guardrails.
Why It Matters
Generative AI is the fastest-adopted consumer technology on record, and it is changing the value of skills faster than institutions are adapting. The evidence shows large, measurable productivity gains, a distribution of risk that falls unevenly on clerical workers, women, and the young, and no visible aggregate effect on measured happiness in either direction. For economists and social scientists, the window 2023 to 2026 is a rare natural experiment, and the report identifies eight research priorities, from algorithmic management and freedom to deskilling and education.
The full report provides 17 interactive ECharts visualisations, 16 downloadable datasets, a methodology document, and a Python replication script.