MAP-AI Explorer · AI adoption in European firms

AI-active firms

—

— —

Firms whose websites show they use, develop or deploy AI

Classified firms

—

— —

Firms matched to classified website content that year

Generative AI

—

— —

AI-active firms working with generative AI and conversational systems

How AI spread through firms

Share of classified firms that are AI-active. Once a firm’s website shows AI, it counts in every later year. Click the chart to pick a year.

AI-active firms by country

Number of AI-active firms each year, stacked by country. Click a bar to pick its year.

Countries in —

Change is in percentage points on the year before. Growth is the share now divided by the share in —.

Country	AI-active share	Change (pp)	AI-active firms	Classified firms	Growth

Regional level

AI adoption across regions in —

Share of each region’s classified firms that are AI-active; darker is higher. The scale stays the same across years, so press Play to watch AI spread. Point at a region for its figures, and click it to find it in the table.

Highest shares

Lowest shares

Boundaries: Eurostat GISCO, NUTS 2024 (© EuroGeographics for the administrative boundaries). The French outermost regions are not on the map; they are in the chart and the table below.

Every region in —

Each dot is a region, placed by its AI-active share and sized by its number of classified firms. The dark tick marks each country’s median region. Press Play to watch the years go by, and click a dot to find it in the table.

All regions in —

Sort by any column, or search by name, code or country.

Industries

Industries from — to —

AI-active share of each industry’s classified firms, ranked. Industries with fewer than — firms have no share.

—
—

Industries by country in —

AI-active share; darker is higher. Hatched cells have fewer than — classified firms. Sort by any column, or search.

Breakdown

Highlight

Technology types of AI-active firms

Share of AI-active firms in each category, by year. A firm can be in several categories, so the shares add up to more than 100%. Click the chart to pick a year.

—
Other technology types

From — to —

Share of AI-active firms: hollow dot —, filled dot —. Click a row to highlight it.

Countries in —

Share of AI-active firms; darker is higher. Click a row to highlight it.

— by industry in —

Share of each NACE section’s AI-active firms; the dashed line is all industries together. Sections with fewer than — AI-active firms have no share.

Survey year

MAP-AI and the Eurostat survey, industry by industry

Each dot is an industry (a NACE section): its share of firms using AI in Eurostat’s survey of enterprises, across, against its share in MAP-AI, up. The table names the letters; its difference is MAP-AI minus the survey, in percentage points.

Correlation r = — across — industries.

Industries
Fitted line
Equal shares

Code	Industry	Survey	MAP-AI	Difference (pp)

Both measures over time

Lines: MAP-AI each year, in the industries the survey covers. Rings: the survey, in the years it asked about AI. Click the chart to pick a year.

MAP-AI
Eurostat survey

How the two measures differ

Who is counted: The survey asks enterprises with 10 or more persons employed. MAP-AI covers firms of every size that have a website, most of them small, and AI is less common in small firms: one reason MAP-AI’s shares are lower.
What counts as AI: The survey asks firms whether they use at least one of eight AI technologies. MAP-AI reads firms’ websites for the use, development or deployment of AI, so a firm that uses AI without saying so online is not counted; once a firm shows AI, it counts in every later year.
Industries: The survey covers NACE sections C to N except finance (K), plus the repair of computers. The comparison uses the sections both report; a few survey figures are confidential.
Years: The survey asked about AI in 2021, 2023 and 2024, but not in 2022. The paper averages the three years.
Several countries: As in the paper, the survey’s share for several countries is the average of the shares they report, and MAP-AI’s pools their firms.
Source: Eurostat, Artificial intelligence by NACE Rev. 2 activity (isoc_eb_ain2): enterprises using at least one AI technology, as a share of enterprises with 10 or more persons employed; the copy used in the paper. © European Union, reused with the source acknowledged.

What the numbers mean

AI-active firm: A firm whose website, read by a large language model, shows that it uses, develops or deploys AI. MAP-AI measures adoption that has happened, not exposure to AI.
Cumulative: Once a firm is classified as AI-active, it counts in every later year. Roles and technology types work the same way.
Classified firms: Firms that could be matched to classified website content in a year: the denominator of every AI-active share.
Roles and technology types: Given only to AI-active firms, and a firm can have several, so the shares of AI-active firms add up to more than 100%.
Several countries together: When you pick two or more countries, this dashboard adds up AI-active firms (or firms in a category) across them and divides by the summed denominator. For industries, all four countries use the repository’s own pooled rows; two or three countries weight each country’s share by its classified firms, leaving out withheld shares.
Heat map: The data place firms only in their small (NUTS 3) region. The heat map estimates each region’s AI-active firms as its share × its classified firms, spreads them around Eurostat’s centre point for the region (more widely for larger regions) and adds them up. It shows where AI-active firms concentrate, not where single firms are.
Withheld shares: A share is withheld when its denominator has fewer than — firms. Here those show as hatched cells or missing dots.

Source and coverage

Source: firm websites from Common Crawl, classified with the MAP-AI method.
Countries: Belgium, France, Germany and Luxembourg.
Years: — to —.
Regions: Eurostat NUTS 2024, levels 1 to 3.
Industries: NACE Rev. 2 sections and divisions, and NAICS 2022 sectors.
MAP-AI x Eurostat: Eurostat’s survey on ICT usage in enterprises (isoc_eb_ain2), as in the paper.

The repository also holds finer industries (NACE groups and classes, NAICS industry groups and industries) and roles and technology types by region, which this dashboard leaves out. Region names and map boundaries come from Eurostat’s NUTS 2024 files (© EuroGeographics for the administrative boundaries). Industry names were written out from the published classifications rather than downloaded, and translated into German and French for this dashboard: check them against Eurostat and the US Census Bureau before citing them.

Cite and reuse

Garbers, Julio and Terry Gregory. “The Diffusion of Artificial Intelligence Across Firms: Evidence from Europe.” IZA Discussion Paper No. 18434, 2026.

Read the paper · Get the data

The data are licensed CC BY 4.0: share and adapt them for any purpose, crediting the paper above.

AI-active shareAnteil KI-aktiver UnternehmenPart d’entreprises actives en IA

AI-active firmsKI-aktive UnternehmenEntreprises actives en IA

Classified firmsKlassifizierte UnternehmenEntreprises classées

Generative AIGenerative KIIA générative

How the two measures differWie sich die beiden Maße unterscheidenEn quoi les deux mesures diffèrent

What the numbers meanWas die Zahlen bedeutenCe que signifient les chiffres

Source and coverageQuelle und AbdeckungSource et couverture

Cite and reuseZitieren und weiterverwendenCiter et réutiliser

AI-active share