UK AI JobsExpected · in use · happened
Deeper cuts · is it just the UK?

No. And the deeper answer is stranger: nobody's statistics name AI.

Every advanced economy is adopting AI at once, so the cross-country spread is a natural experiment. Two things travel across it. First, real usage inverts with income: where AI is scarce it automates, where it's everywhere it augments. Second, and this one runs deeper, the realised-AI-displacement cell is empty in every country's official data. That's an institutional choice, not a data limit.The UK isn't uniquely hit. It's the loudest reader of a silence that’s everywhere.

Measured · Anthropic Economic Index · 121 countries

Where AI is barely used, it automates. Where it's everywhere, it augments.

One measured cross-country pattern, no guesswork. In the third of countries where AI is least adopted (Tanzania, Uganda, Mozambique) usage skews toward automation (47.6% augmentation). In the most-adopted third (Singapore, Switzerland, Australia) it skews toward augmentation (52.7%). Adoption tracks national income, so this is the automation-vs-income inversion, drawn entirely from real usage. The UK sits high on both: 55.1% augmentation at 3.38× adoption.

Measured: AEI, real Claude.ai usage, Apr+May 2026, CC-BY. Claude users only, so adoption reflects Claude's footprint (a proxy for AI adoption, itself a proxy for income), and “augment vs automate” is how AI is used, not a count of jobs lost. Method →

The same empty cell, three ways

It isn't that the data is missing. It's that no one points it at AI.

United Kingdom

HMRC Datalab and ONS secure microdata are open and heavily used, by the exact institutes that never point them at AI attribution. The flagship release: 0 mentions of AI in 37 pages.

Germany

Owns arguably the world's best linked employer–employee data (IAB). It routes 37 AI publications into adoption and potential registers, and zero into realised AI-attributed displacement.

Canada

StatCan runs the official stats, builds the linked microdata, and operates the academic-access network. One roof, one custodian. The realised cell is still empty.

The one study that didmeasure realised displacement, Stanford's ~16% drop for AI-exposed under-25s, ran on private payroll data, not government statistics. That's exactly where the measurement gate predicts such a number is allowed to live: a neutral, well-resourced team with the right data finds it; the official ledgers, by mandate, don't look.

Read levels with care. Youth-unemployment figures are notcomparable in level across these rows. Countries use different age bands (Korea 15–29, most 15–24) and student-inclusion rules (Sweden's 24.4% is ~79% full-time students; the UK is shown on the incl-students MGWY=16.2% basis). The comparison is valid on direction and archetype, not on a strict ranking. AI job-posting share is a single-source Indeed Hiringlab Feb-2026 reading for four economies; the rest are shown as "no data" rather than guessed.

The absorber map, and why the UK would show it first.

UK highlighted

Same AI wave, different shock absorbers. Strong absorbers (dual-vocational, demographic-pull) hold youth unemployment down; the deteriorating ones, the UK among them, let it climb. And the UK leads its cohort on AI-adoption intensity, so if a realised youth signal is coming, this is the labour system where it should surface first.

Sweden
24.4%
Spain
23.7%
France
21.5%
Italy
16.9%
United Kingdomyou
16.2%
Canada
14.3%
Ireland
10.2%
Australia
10.1%
Netherlands
8.7%
Germany
7.5%
South Korea
7.4%
Japan
3.8%
United States
n/a
AI job-posting share · where measured
Indeed Hiringlab · Feb 2026 · 4 economies
United Kingdom
7.5%
United States
4.9%
Germany
4.1%
France
3.4%

Sources and per-figure provenance: see public/data/international_sources.md. Every figure traces to a primary release; missing cells are shown as "no data", never imputed.