Data Descriptor
GAIA: A Global AI Adoption Index Combining Observed Usage, Theoretical Exposure, and Pre-GPT Baselines Across 923 Occupations and 138 Countries
International Finance Corporation, World Bank Group
Abstract. Research on AI and work has fragmented into three measurement traditions that rarely speak to each other: observed usage from AI platform logs, theoretical exposure from task-level ratings, and pre-GPT baselines capturing prior automation and digital adoption. GAIA — the Global AI Adoption Index — harmonizes all three onto a single crosswalk, covering 923 O*NET occupations and 138 countries. This descriptor documents the index construction, variables, and coverage, and shows that observed usage and theoretical exposure diverge substantially, implying that the choice of measure materially changes conclusions about who is exposed to AI.
Figure 1 · Distribution of GAIA-E across 923 occupations
Histogram of the GAIA-E exposure score (0–1). Dashed line marks the mean.
Source: data/gaia_occupations.csv · GAIA-E (gaia_e)
Figure 2 · OpenAI exposure vs. SML score
Each dot is an occupation, colored by major occupation group. Two theoretical measures, weakly aligned.
Source: data/gaia_occupations.csv · OpenAI Eloundou β (dv_rating_beta) vs. SML (sml_score)
Figure 3 · Top 20 and bottom 20 occupations by GAIA-E
Most-exposed (teal) and least-exposed (rose) occupations by GAIA-E.
Source: data/gaia_occupations.csv · GAIA-E (gaia_e)
Figure 4 · Country panel — usage vs. work-use, sized by GAIA-A
Bubble = country. x: global usage share · y: work use-case share · size: GAIA-A adoption score · color: income group.
Source: data/gaia_countries.csv · usage_pct_global, uc_work, gaia_a, income_group
Figure 5 · Mean GAIA-E by major occupation group
Average GAIA-E exposure within each major occupation group, ranked highest to lowest.
Source: data/gaia_occupations.csv · GAIA-E (gaia_e) by group
Key findings
—
Occupations scored on a common O*NET crosswalk
—
Countries with harmonized usage & adoption measures
—
Mean GAIA-E across occupations (0–1 scale)
Data access
GAIA data is currently available by request. The full occupation- and country-level datasets, the Zenodo deposit, and the variable codebook are released to approved users via the waitlist.