Anthropic publishes a per-capita measure of how much a country uses its AI assistant. It divides a country's share of global usage by its share of the world's working-age population. A value of 1.0 means the country uses AI exactly in proportion to its size.
In May 2026 the UAE reads 2.84. Saudi Arabia reads 0.93.
The UAE ranks 23rd of the 121 countries with a published value. Saudi Arabia ranks 64th. One is using AI nearly three times as intensively as its population would predict; the other is using it slightly less than its population would predict. They are usually sold to investors as one market.
Every figure in this article is our own computation from Anthropic's open data file for May 2026. Anthropic has not published these country cuts.
Six countries, one label
| Country | Usage index | Rank (of 121) | Coursework | Work | Personal |
|---|---|---|---|---|---|
| UAE | 2.84 | 23 | 9.09% | 54.63% | 36.28% |
| Qatar | 1.66 | 41 | 11.65% | 50.85% | 37.51% |
| Bahrain | 1.50 | 44 | 25.11% | 39.06% | 35.83% |
| Saudi Arabia | 0.93 | 64 | 16.59% | 47.26% | 36.16% |
| Kuwait | 0.92 | 65 | 15.11% | 40.06% | 44.83% |
| Oman | 0.73 | 70 | 25.03% | 43.16% | 31.81% |
| US, for reference | 3.87 | 12 | 8.94% | 41.32% | 49.74% |
The level splits the six in half. The UAE, Qatar and Bahrain sit above population parity; Saudi Arabia, Kuwait and Oman sit below it. That is a wider internal spread than the gap between Saudi Arabia and the regional average.
The composition splits them differently, and more usefully.
The UAE is a work market. More than half of all conversations there are classified as work, the highest work share in the table and above the United States. Coursework is 9.09%, essentially the American level.
Bahrain and Oman are learning markets. Both sit at roughly a quarter of all conversations classified as coursework — around two and a half times the UAE's share. Bahrain's usage index is above parity and Oman's is well below it, which means the coursework pattern is not simply a function of how much AI a country uses.
Saudi Arabia is in between, and moving. Coursework is 16.59%, nearly double the UAE's. On request topics, Education & Learning is Saudi Arabia's second largest category at 13.82%, behind Content Creation at 22.94%. In the UAE, Education & Learning is fourth at 9.99% — the same rank it holds in the United States at 9.28%.
The teaching signal points the same way. The share of tasks that map to the "Training and Teaching Others" work activity is 3.54% in Saudi Arabia against 2.65% in the UAE and 2.22% in the United States. On the broader occupational cut, tasks mapping to the Educational Instruction and Library occupational group run 15.10% in Oman, 13.60% in Bahrain, 13.39% in Kuwait, 12.82% in Qatar, 11.70% in Saudi Arabia and 10.70% in the UAE. The UAE is the least education-weighted market in the Gulf and the most work-weighted.
One thing the data does not support
The September 2025 edition of the index reported that low-adoption regions skewed toward automation rather than collaboration — that where AI is scarce, people hand it whole jobs rather than working alongside it.
That does not hold for the Gulf in the May 2026 data. Augmentation shares cluster tightly: UAE 49.72%, Qatar 51.92%, Bahrain 52.50%, Saudi Arabia 51.39%, Kuwait 51.14%, Oman 49.92%, against the United States at 50.41%. The lowest-adoption markets in the table are not the most automation-heavy. The UAE, the highest-adoption market, has the lowest augmentation share of the six.
Whatever separates these countries, it is not how they collaborate with the model. It is what they bring to it.
Three cautions, because this metric is easy to misuse
The values are renormalised between releases. Israel was reported at 7.0 in September 2025. In the May 2026 file it reads 3.05 and ranks 21st, with Australia now highest at 6.40. Nothing about Israel changed by half. The scale did. Values from different releases are not a time series, and anyone charting them as one is drawing a picture of a methodology change.
The denominator is population, not workforce composition. The index divides usage by working-age residents. It does not ask what those residents do. A country whose resident population skews heavily toward white-collar and professional work will read high for reasons that have nothing to do with policy, procurement or school systems. That is a meaningful part of the UAE-Saudi gap and the data cannot tell you how much.
This is one vendor's users. Claude is not the only assistant in any of these markets, and it is almost certainly not the most used. Suppressed cells mean "not published", not zero. Treat the composition as a signal about a self-selected population, not a national statistic.
Anthropic's own March 2026 edition sharpens the last point in a way that matters more than the country ranks: geographic concentration is getting worse, with the top 20 countries now accounting for 48% of per-capita usage against 45% previously, and the estimated time for the rest of the world to reach US per-capita adoption revised from two-to-five years out to five-to-nine.
What it means for a go-to-market plan
A single "GCC strategy" is two or three different products wearing one label.
The UAE looks like a Western enterprise market on this data: work-dominated, above parity, low coursework. Selling seats to organisations is a plausible motion there. Saudi Arabia, Oman and Bahrain look like learning markets, where the demand showing up is coursework and teaching, and where a per-seat enterprise price will meet a customer who is not the one generating the usage.
The venture data points the same direction and harder. In the first half of 2026, UAE startups raised $895 million across 79 deals while Saudi Arabia raised $219 million across 72 deals, a 74% fall. Similar deal counts, four times the capital. A founder building for Saudi Arabia is building for the market with the education demand and the smaller cheques.
Both governments made AI a mandatory school subject from the 2025-26 academic year — the UAE from reception through grade 12, Saudi Arabia across all levels of public education with SDAIA involved in delivery. That is the same policy in two countries whose actual AI usage looks nothing alike, which is a useful reminder that a mandate is an input, not an outcome.
What a founder should do about it
Price and sequence per country, not per region. The same product at the same price into the UAE and Oman is two different businesses, and only one of them will work.
Follow the composition, not the rank. A low usage index with a high coursework share is a market with unmet learning demand and no incumbent. That is a better opening than a saturated work-dominated market where you are competing with a general assistant on its home ground.
Stop quoting a MENA average. There isn't one. The ILO's exposure work makes the same point from the labour side, putting Arab States at 5.8% of youth jobs in the most exposed categories against 14.3% in high-income countries, while the Gulf itself straddles both income brackets. Any single regional number is an average across markets that behave nothing alike.
Compute your own market's cut. The file is open, it is CC-BY, and the analysis in this article took an afternoon. Whatever country you are selling into, the composition of its AI usage is knowable, and almost nobody has bothered to look.
Two point eight four and nought point nine three. Same region, same month, same product, same metric.