📐 Methodology

How often the world reads about someone or somewhere: the number of times a person's, a country's or a metro's Wikipedia article was read, in every language edition, in a month. Attention is not approval and it is not power. A war, an election or a World Cup debut can multiply a month several times over, and that is the point: this is what people looked up.

Wikimedia's pageview_complete dumps, user traffic, one file a month, each article's daily views summed over desktop, mobile web and the apps. Each item's title in every language comes from its Wikidata sitelinks. The dumps are kept on our own machine, so a month can always be recounted.

People are our own Nowhere 100, matched to Wikidata through their English article. A name that leads to a disambiguation page is left out rather than guessed. Countries are read through the country's own article, not its cities or history pages. Metros are read through the primary city's article, accepted only when that article's coordinates sit within 60 km (37 mi) of the metro, so a city that shares a name with a better-known one is not credited with its readers.

Totals: 94 people, 233 countries and territories, 3,929 metros.

Months on record run from September 2025 to September 2026. The index shows one month by default, the last 12 months (October 2025 to September 2026), or any range of up to 12. The attention score is 0 to 100 within each group: 100 times the entry's readers over the period divided by the most-read entry in the same group. People, countries and metros are never scored against each other.

The user dumps still carry automated reading that Wikimedia did not catch. Most of it follows human interest, such as an assistant looking something up for a person, and it is kept as a sign of that interest. Two shapes are not interest, and they are removed day by day, each article cut back to its own normal level:

• A flat sweep. Hundreds of articles in every language gaining the same steady amount every day for weeks. Between 31 October and early December 2025 one program added about 80 to 90 views a day to every language of some 32,700 place articles. A rise is treated as a sweep only when its daily excess is flat and the same shape appears on at least 500 articles in the month: real interest is never flat.
• A twin loop. One program reading two unrelated articles at the same count every day: the English articles on Durrës and Utrecht, 25,000 to 45,000 views a day each until 21 November 2025, then a few hundred. News lifts related articles together but never to identical numbers, so pairs such as Kampala and Zohran Mamdani in his election week are kept.

Removed in all: 90.9M views (September 2025 1.7M, October 2025 1.9M, November 2025 87.1M, December 2025 121K).

Wikipedia is blocked on the mainland, so Chinese cities read far below their size. Until a Chinese source can be counted, mainland metros carry an estimate built on Taiwan, marked est. and ranked on it:

• Taiwan's cities are read almost only from inside Taiwan, in Chinese: Wikimedia's pageviews-by-country data puts 96 to 100 percent of those views in Taiwan. That gives a measured rate of Chinese-language reading per resident, about 0.0032 a month.
• That rate times the population of China's metros, scaled by internet use (80 percent in China, CNNIC, against 97 in Taiwan), is the missing domestic reading.
• It is shared among Chinese metros by the Chinese-language reading each already gets.

Over the last 12 months that multiplies mainland metros' measured readers by about 1.83. Rank within China is the reliable part; the level is a placeholder.

The month so far is the sum of Wikimedia's daily dumps, the same files the monthly figure is built from, updated every Sunday and at month end. It is provisional: the sweep tests need a whole month and its neighbours, so they run on the monthly index, not on the weekly view.

• Language is a proxy for market, not country: English-language reading comes from everywhere.
• A metro named for one city counts that city's readers only, not its suburbs' own articles.
• The China estimate rests on one benchmark; it is shown as an estimate, never as a measurement.
• Automated reading that does not look like a sweep is counted, by design.

Cite this: Who the World Reads About a1.202609, Citizen of Nowhere, as of 27 Oct 2026. https://rankings.citizenofnowhere.org/attention. Reuse with credit under CC BY 4.0.