Wealthier countries don’t just adopt AI in name — they use it far more than everyone else
As AI adoption accelerates worldwide, a new question matters as much as whether a country can access AI: how intensively its population actually uses it. The Anthropic AI Usage Index, a 2026 measure of AI activity intensity by country, shows usage is concentrated sharply among the wealthiest economies — echoing the same income divide that shapes so much else about how nations experience new technology.
Across the ten wealthiest of the 20 countries in this dataset (GDP per capita above $35,000), the average AI Usage Index reads 3.55 — more than four times the 0.85 average among the ten lower-income economies. GDP per capita and AI Usage Index correlate at r=0.87 across all 20 countries, one of the cleanest relationships in this dataset: the US, Canada and Australia lead on both measures; Nigeria, Ghana and India lag on both.
View data
| Ratio | |
|---|---|
| GDP per capita (USD) | 6.14x |
| AI Usage Index | 4.16x |
AI usage tracks national income closely, but not perfectly: developed economies use AI about 4.2x more than emerging markets — a narrower gap than their 6.1x GDP advantage, meaning some countries adopt AI faster than income alone would predict.
Mexico is the clearest example of an economy lagging its own income bracket: despite the second-highest GDP per capita among the ten emerging economies here ($15,780), its AI Usage Index of just 0.63 trails lower-income peers like Argentina (1.48, GDP $14,360) and Colombia (1.18, GDP $10,100) by a wide margin.
South Korea shows the mirror case: with GDP per capita of $37,410 — the second-lowest among the ten developed economies, just above Japan — it still posts the fifth-highest AI Usage Index in the entire sample at 3.78, ahead of wealthier peers like Germany ($65,300, AI usage 2.40) and Sweden ($70,680, AI usage 2.98). Mexico and South Korea sit at opposite extremes of the same phenomenon: income sets a country’s AI-usage ceiling only loosely, and something else — infrastructure, digital culture, policy — determines how close a country actually gets to it.
The mechanism is plausible: AI usage depends on digital infrastructure, cloud computing access and a services economy large enough to generate obvious use cases — all of which correlate with GDP per capita without being identical to it.
Population adds a different kind of nuance. Today’s AI usage gap is real, but it is a gap in intensity, not in the size of the population that could eventually close it: most people live in the markets where the AI Usage Index has the most room to grow, led by population giants like India (1.48 billion) and Indonesia (288 million).
AI adoption is accelerating worldwide, creating new opportunities for emerging markets to drive long-term economic transformation. However, the spread of AI across markets is progressing unevenly, held back by weak infrastructure, limited access to quality data, shortages of specialized talent and institutions that are not yet fully prepared to support adoption.
It’s usage itself — not income — that best predicts how a country’s workforce feels about AI and jobs.
The more a country actually uses AI, the more its workforce expects AI to cost jobs
Averaged across the ten highest-usage countries in this dataset, 60.7% of adults expect AI to mean fewer jobs. Among the ten lowest-usage countries, only 41.6% do — and the AI Usage Index correlates with that pessimism at r=0.88 across all 20 countries, a stronger relationship than GDP per capita’s own r=0.74. Usage, not income, is the more precise predictor of how a workforce feels about AI’s impact on jobs.
The countries with the highest AI Usage Index scores — Australia (6.40), Canada (4.13), France (3.97) and the US (3.87) — all also rank among the most pessimistic about job losses. The lowest-usage countries — Nigeria (0.24), India (0.30), Ghana (0.38) and Indonesia (0.46) — are also the most optimistic. The pattern holds almost without exception at the extremes.
Peru is the sample’s clearest exception: despite a modest AI Usage Index of 1.05 — lower than Chile, Argentina or Colombia — 57% of Peruvians expect AI to mean fewer jobs, the most pessimistic reading of any emerging economy.
That combination — high anxiety without correspondingly high usage — looks less like a reaction to AI already reshaping Peruvian workplaces and more like anticipatory concern: a workforce bracing for a wave of adoption it expects to arrive rather than one it has already lived through.
Mexico shows the opposite pattern: its AI Usage Index of 0.63 is among the lowest in the emerging tier, and its 44% pessimism sits near the middle of that tier too — sentiment that, for now, still roughly tracks actual usage rather than running ahead of it.
Together, Peru and Mexico frame the two ways an emerging economy’s sentiment can diverge from its usage level: anxiety can arrive early, before adoption, or it can lag until usage itself catches up. Distinguishing the two matters for how — and how urgently — an organization should communicate about AI in each market.
Track AI usage, not just income, to see where job-loss anxiety is headed next
- 01Use AI adoption data as a leading indicator of workforce sentiment, not just a business metricBecause the AI Usage Index predicts job-loss pessimism more precisely than GDP per capita, a market’s usage trajectory is a more useful early-warning signal than its income trajectory. Organizations planning AI-adoption communication or reskilling programs should track usage growth in a market before sentiment shifts, not after.
- 02Don’t assume income determines a market’s readiness for AI messaging — usage doesChile and Argentina already show AI Usage Index levels approaching the developed-market floor, well ahead of their income peers; a message calibrated only to their GDP tier would under-prepare for the anxiety already building there.
- 03Watch Peru and Mexico as the two clearest tests of this frameworkPeru’s pessimism already runs ahead of its usage level — worth confirming with qualitative research whether that reflects genuine anticipatory concern or a measurement artifact. Mexico’s usage remains genuinely low; if it rises toward its income-tier peers, sentiment should be expected to follow, not lead.
None of this settles whether AI will, in fact, produce fewer jobs or more — that verdict differs by industry, occupation and time horizon. What the data does establish is that usage, more than income, is the variable to watch: it explains more of the variation in how workforces feel about AI than wealth alone ever did.
