There is only one major topic in which Italy ranks among the world’s top five countries by its share of Claude conversations, and it is regulatory compliance. Out of the 96 countries for which Anthropic publishes the figure, we are fifth, with 1.6% of conversations against a 1.1% world average, and in none of the other nineteen major topics do we do better than nineteenth. Careful, though: it is a weight, not a volume, because per person a Swiss user uses Claude for compliance twice as much as we do.
I took the May 2026 snapshot, 121 countries, and looked at it through the eyes of an Italian entrepreneur, lining up the earlier releases since August 2025 as well. The short answer is that we use Claude little, and when we open it, paperwork weighs more than elsewhere. In the interest of transparency: I have just graduated studying with AI, and I wrote the report behind this piece with Claude too.
What do Italians use Claude for?
In broad terms, for the same things as everyone else. 43% of Italian conversations are about work, 41% about personal life and 16% about study, just like the world average, and the most frequent topics are the same too.
Source: Anthropic Economic Index, May 2026 data.
Source: Anthropic Economic Index, May 2026 data.
The Italian accent shows up next to seven countries similar to ours, the same ones as in the charts below. We come first in regulatory compliance and in document processing, which accounts for 5.3% of our requests against 4.3% for the world average. I picture it like this: pulling data out of invoices, turning a PDF into a table, summarising a public tender, the office work that nobody in a small business has time for and everybody has to do. On the business side it is the other way round: we come last in sales (0.8% against 1.1% for the world) and second to last in business processes (3.5% against 4.7%), ahead only of France.
Compliance and regulation, %
Document processing, %
Business processes, %
Sales, %
Source: Anthropic Economic Index, May 2026 data.
What we ask Claude to produce has the same accent. Explanations and answers make up 18.8% of conversations, 13% above the world average. Plans and strategies are 12% below, and among the eight we come last. We use Claude more like a teacher than a strategy consultant: strategy, it seems, we prefer to work out on our own, ideally over an espresso at the bar.
Source: Anthropic Economic Index, May 2026 data.
Before turning this into a theory of Italian bureaucracy, I looked for the data that would disprove it. Across all the countries with data, document processing weighs more where Claude is used little, business processes and plans and strategies where it is used a lot: part of our profile is simply typical of countries that use it little. Taking that into account, document processing is still 16% above what you would expect and business processes 16% below, while on plans and strategies the gap shrinks to 7%. Compliance, on the other hand, weighs more where Claude is used more, and we sit more than 50% above what our level of usage would predict. That is the truly Italian trait.
How much is Claude used in Italy compared with other countries?
Anthropic’s index divides a country’s share of conversations by its share of the population aged 15 to 64, and it equals 1 when a country uses Claude exactly as much as the world average. Italy scores 1.7, the 39th value out of 121. It sounds like a good grade, until you look at the classmates sitting next to you. Among the eight countries compared we come last, and in the European Union we are 21st of 27, with only Greece and five Eastern European countries behind us. Switzerland uses it almost three times as much as we do, Portugal almost twice as much.
Source: Anthropic Economic Index, May 2026 data. Index computed on the 15-64 population.
Income is not enough to explain it. According to Anthropic, a GDP per capita 1% higher is associated, on average, with usage about 0.7% higher, yet Portugal has a GDP per capita about a quarter lower than ours ($32,000 against $43,000 in 2025, according to the International Monetary Fund) and a population almost as old as ours. At the bottom of the table sit Europe’s two great manufacturing economies, Italy and Germany: a hunch, not proof, but I take it home with me.
Which Italian regions use Claude the most?
For regions, Anthropic only publishes each one’s share of national usage. I divided it by each region’s share of the 15-64 population (ISTAT, the Italian statistics office, estimate at 1 January 2026), so that 1 means in line with Italy. It is my own index, and it cannot be compared with the country index.
Nicola Giunchi's elaboration on Anthropic Economic Index (May 2026) and ISTAT data. Boundaries: openpolis.
Nicola Giunchi's elaboration on Anthropic Economic Index (May 2026) and ISTAT data (population at 1 January 2026, estimate). Calabria, Basilicata, Molise and Aosta Valley: not published.
Lazio and Lombardy, the regions of Rome and Milan, account for 57% of Italian usage with 27% of the working-age population. The figure that stopped me was Veneto: more inhabitants than Piedmont and 35% less usage, at 0.57 of the national average. Together with Trentino-Alto Adige (0.61) and Friuli-Venezia Giulia (0.40), the whole Triveneto, the three north-eastern regions, sits below two thirds of the average. Emilia-Romagna, where I live, is at 0.95. Right on the average: for someone from Romagna, almost a personal insult.
One caveat. Regions are assigned by IP address, and corporate networks that reach the internet through Milan or Rome can inflate Lombardy and Lazio. But the pattern seems to me too sharp to be just an optical effect. And it is not a May accident: in all four periods with regional data, Lazio and Lombardy account for between 55% and 57% of usage.
Is Claude usage in Italy growing?
Anthropic does not publish a time series, so I rebuilt the trend from four public releases, from August 2025 to May 2026, recalculating the index with the same formula every time. Until February, however, each release covered one sample week of chat only, and in February the Max plan enters the sample. From April it covers the full month and includes Cowork: across the change, only countries can be compared with each other.
Nicola Giunchi's elaboration on Anthropic Economic Index data, releases of September 2025, January, March and June 2026. April and May are recalculated, not the official values: the recalculation gives 1.56 and 1.69 for Italy against the official 1.56 and 1.70 (gaps of 0.01 to 0.03 for the other countries). August and November: Free and Pro plans; from February, Max too.
From August to February Italy rises from 1.75 to 2.14 (+22%), like Germany, while France grows by 54% and Spain by 50%. Spain overtakes us in February and by May it is 59% ahead. At the change of method, between February and April, the Italian index falls by 27%, and no country in the comparison does as badly.
Nicola Giunchi's elaboration on Anthropic Economic Index data, March and June 2026 releases. Among all 113 countries with both values, Italy is the 13th worst (median +1.1%; just ahead of us Poland, −28%; the steepest fall is Iraq's, −47%).
The data cannot say why. Perhaps the February sample week was unusual, perhaps the new method weighs differently from country to country. Or perhaps some Italian users have moved to other assistants, the most worrying hypothesis and one these numbers can neither confirm nor rule out. From April to May, at least, things pick up: +8.5%, in line with Spain and Portugal.
| Country | Aug 25 | Nov 25 | Feb 26 | Apr 26 | May 26 |
|---|---|---|---|---|---|
| Italy | 1.75 | 1.92 | 2.14 | 1.56 | 1.69 |
| Italy, rank | 35th of 115 | 32nd of 117 | 31st of 116 | 40th of 114 | 39th of 121 |
| Spain | 1.50 | 1.93 | 2.24 | 2.43 | 2.69 |
| France | 2.31 | 3.16 | 3.55 | 3.87 | 3.97 |
| Germany | 2.18 | 2.12 | 2.66 | 2.36 | 2.40 |
| Portugal | 2.39 | 2.64 | 3.03 | 3.06 | 3.34 |
| United Kingdom | 3.16 | 3.08 | 3.77 | 3.39 | 3.35 |
| Switzerland | 3.34 | 3.81 | 4.93 | 4.71 | 5.04 |
| United States | 4.30 | 4.38 | 4.66 | 4.25 | 3.87 |
What do these numbers say, and what don’t they?
They measure Claude, not AI in general, and Italy may simply prefer ChatGPT or Gemini. As Anthropic itself warns, they say nothing about the effects on employment. Nor do they say that paperwork is what holds us back, only that it weighs more here than elsewhere.
As an entrepreneur, the thing that stays with me is the country of paperwork. My reading is this: everywhere, people start with the most tedious work to delegate, and the risk for Italy is staying there. That means using AI to defend ourselves from bureaucracy instead of using it to sell, plan and decide. In a small business, freeing up those office hours is already a gain. The next step is to spend them on the business. Outside Lazio and Lombardy almost all of Italy still uses it little, so whoever starts now starts ahead of the competitor next door. And without yet another subscription, because software can now be built in-house. What I build with it myself, I describe on this page. The full report, with the charts that do not fit here, I will send you by email (in Italian).
Sources: Anthropic Economic Index, May 2026 data, CC BY 4.0 licence; public releases of September 2025, January, March and June 2026 (full topic data, not just the top ten) with their documentation on Hugging Face; Anthropic, Economic Index report, September 2025 on the geography of adoption; ISTAT, resident population by age at 1 January 2026 (demo.istat.it); International Monetary Fund, World Economic Outlook, April 2026 (GDP per capita 2025 in US dollars); Eurostat, demographic indicators (median age); regional boundaries: openpolis, geojson-italy. The regional index, the time series and the comparisons with the expected values are my own elaborations, made with Claude on the public Hugging Face files and through the Anthropic Economic Index connector; independent analysis, not affiliated with Anthropic.