Wikipedia may look like a collection of articles, but beneath the familiar pages lies something far more interesting: a gigantic network. The English Wikipedia contains more than 65 million pages, including approximately 7 million articles. Connecting them is an enormous network of 1.63 billion unique internal page-to-page links (wikilinks).
Every blue link creates a connection between two pieces of knowledge. Add them all together, and Wikipedia begins to resemble a map of how people organize history, science, culture, technology, and everyday life. But which articles occupy the most important positions in that map?
Watch the full video:
Beyond Page Views: Measuring Wikipedia by Links
The obvious way to measure an article’s popularity is to count how many people read it. That tells us what attracts attention at a particular moment. Link analysis asks a different question: how many other articles depend on this page as a reference point?
An article with thousands of incoming links is structurally important. It acts as a hub—a destination repeatedly used to explain, classify, or connect other topics. At first glance, we might expect broad subjects such as “United States”, “World War II” or “Human” to dominate. Yet Wikipedia’s internal structure is shaped not only by famous historical topics, but also by its technical standards, citation systems, templates, identifiers, and editorial conventions. To discover the true leaders, the entire network has to be processed.
Why Reading the Wikitext Is Not Enough
A simple script could download Wikipedia pages and search their source code (wikitext) for wikilinks. In theory, that sounds straightforward. In practice, it would produce an incomplete and potentially misleading result. Wikipedia’s source text is highly irregular. Links can appear inside infoboxes, citation templates, navigation boxes, formatting macros, and modules connected to Wikidata. Some are typed directly by editors, while others are generated automatically by the software.
Then there are redirects. For example, a link to “USA” ultimately leads to “United States”. Without resolving that redirect, the importance of one topic would be split between multiple names.
Four Database Tables, One Giant Graph
The solution was to work directly with Wikipedia’s raw SQL database dumps from May 2026. The analysis combined four essential parts of Wikipedia’s infrastructure.
- The page registry identifies every page and distinguishes encyclopedia articles from user pages, discussions, categories, and redirects.
- The link target dictionary converts internal numerical identifiers into readable article titles.
- The redirect map reveals where abbreviations, aliases, and alternative names ultimately lead.
- The page-link graph records billions of relationships between source pages and their destinations.
Together, these files reconstruct the encyclopedia as a network rather than a stack of documents.
One Article, One Vote
Counting links also requires discipline. If an article links to the same destination five times, those five links should not be treated as five independent endorsements. The relationship between the two articles exists only once. For that reason, each source article was allowed to contribute a maximum of one citation to a target article.
The same rule applied to redirects. If one page linked to both “USA” and “United States”, the redirect was resolved first. The system then recognized that both links pointed to the same destination and counted only one connection. This process prevents repeated links and alternative names from artificially inflating the ranking.
The Template Question
Some of Wikipedia’s most common links are generated through templates, infoboxes, references, and navigation systems. Should they count? There is no simple boundary between a “real” editorial link and an automated one. A manually inserted link may later be moved into a template. An infobox may generate a connection that is central to how readers understand the subject. A citation template may create standardized links to identifiers such as DOI or ISBN.
Removing all template-generated links would therefore require a long list of subjective decisions about which parts of Wikipedia are meaningful. The analysis takes a broader approach. It uses the complete internal graph recorded by MediaWiki – the structure Wikipedia itself recognizes after processing its pages.
Finding the Hubs of Human Knowledge
The resulting ranking does more than identify widely referenced topics. It also reveals how Wikipedia works as a technical system. The most connected pages reflect editorial habits, citation standards, metadata, templates, redirects, and automated modules. In other words, the ranking captures both human knowledge and the machinery used to organize it.
That is what makes the analysis so revealing. Wikipedia is not merely a website containing millions of articles. It is a living information network, built from billions of relationships – and those relationships show which concepts hold the entire structure together.
The ranking is available in this video. The underlying datasets can also be explored on Hugging Face and Kaggle.
The most recent ranking of the most connected Wikipedia articles across different language editions can be found in the WikiRank Citation Index.