Method and sources

Three categories, one rule

Each job is placed using two measures: its exposure to generative AI and its employment outlook.

• New jobs: jobs born from AI or made necessary by it (see below).
• Declining: employment is falling. When a source cites the cause, the job page quotes it with that source. The site never assigns a cause on its own.
• Changing: employment holds or grows, and AI is changing the work. Each page gives the level of exposure: high or very high, the job changes deeply; low or moderate (manual, care, energy jobs…), it changes little.

Rule: in the US version, a job is "Declining" when the BLS projects employment to fall by 1% or more by 2035 (a smaller dip counts as stagnation). Otherwise it is "Changing", and its page says whether its AI exposure is high or low. Exposure is the higher of the two ratings, the ILO's or the BLS's; each job page says which one is used.

Job ads that mention AI

The home page tracks, country by country, the share of online job ads that mention AI (artificial intelligence, machine learning, data science, generative AI…). Source: Indeed Hiring Lab's AI Tracker, open data published monthly; the site shows monthly averages. This figure measures employer demand for AI-related skills across all jobs; it does not say whether AI creates or destroys jobs. Spain is not covered.

Ranking order

"Declining": from the job losing the most positions to the one losing the fewest, based on BLS projections for the United States to 2035. For pages built from official classifications, this volume is an estimate: each US occupation is shared out among the occupations linked to it, with no double counting. "Changing": by employment outlook, then hiring difficulty.

The risk index

A score out of 5, readable at a glance. It does not judge the people who do the job: it summarises two public measures.

Five gauges on every job page

The English version shows data for the United States. When a figure is missing, the gauge says so and names the source used instead.

Is the United States ahead?

US companies often adopt AI earlier, and US projections already factor in its effects. One hypothesis is that the US labour market is a step ahead. The site does not take it for granted; it makes it testable. The French, German, Spanish and Italian versions show their national figure next to the US figure, and every week our review compares trends on both sides of the Atlantic, job by job.

Almost 400 job pages

The site covers the occupations of the International Standard Classification of Occupations (ISCO-08) for which the ILO measures AI exposure, excluding residual categories. About 140 pages are written job by job; the others are built from official classifications: titles and descriptions from the European ESCO classification, and US figures linked through the official ESCO–O*NET crosswalk.

Each page lists the job titles it covers, and search recognises all of them.

New jobs born from AI

The site's first category: new jobs born from AI or made necessary by it (for example MLOps engineer, prompt engineer). They have no official statistics yet, so they cannot be ranked like the others. A job enters only if a source on our list explicitly links it to AI and the team has approved it. It is ranked by four kinds of evidence: reference report, official statistics, recent job ads, recognised training. Jobs growing fast for other reasons, such as the green transition, appear under "Changing".

From figure to page

An AI agent monitors a closed list of official sources, extracts each figure with the exact sentence that contains it and proposes an update. A person approves every week before anything is published.

Advertising

The site shows no advertising for now. It may later be funded by advertising, under these rules. At most one ad slot per screen, always labelled "Advertisement", no pop-ups or autoplay video. No advertiser influences a rank, a gauge or a recommendation.