Understand the source

Wikipedia analysis suite

Choose the view that answers your question: article text, language editions, referenced files, direct categories, outgoing links or daily attention.

Six tools, six different kinds of evidence

The plain text extractor retrieves an article’s readable text and carries source attribution into its exports. It does not turn text into a complete media inventory, count all related topics or establish how many people read the article. Those questions need different source records.

The image and file extractor follows referenced file titles to returned metadata, including individual source pages and available credit notices. Referenced files can include icons, maps and other assets beyond photographs. An accessible original-file URL is not, by itself, permission to reuse the file.

Connections are not all the same relationship

The internal link grapher shows a bounded one-hop view of outgoing article links. The central node and draggable surrounding nodes help organize those records visually; their positions do not measure importance, agreement or the strength of a connection.

The category explorer instead lists direct category members and separates pages, subcategories and files. A subcategory is one returned member rather than a count of everything below it. Categories can have multiple parents, so a category structure is not necessarily a simple tree.

The language edition finder provides links to independently maintained articles in other editions. A linked title is not a guarantee of a word-for-word translation, matching sources or equal completeness. Inspect the destination edition for the actual content you intend to use.

Attention over time needs explicit filters

The daily pageview analyzer requests thirty or sixty complete UTC dates ending yesterday. Access and agent filters remain visible in the table. Pageviews are request counts, not unique readers, edit frequency, article quality or a prediction of future interest.

A missing daily record is not silently replaced with zero. A partial response stays labeled partial, and the average uses the dates actually returned. A large endpoint percentage change can depend on an unusually low starting date; it is not the same as a fitted trend or an explanation for a traffic spike.

A practical research sequence

Start by identifying the exact article and edition. If you need a quotation or passage, use the text extractor and check the article’s source and history. If you need a picture, open its individual file page and review the returned author and license notices. Use the graph or category view to find further reading without treating either as a complete map of the subject.

For a comparison over time, keep the article title, date window, access filter and agent filter together with the exported view counts. A renamed article or redirect can have a separate historical series. Moving to another language edition is also a new source selection, not merely a change in display language.

What happens when you fetch data?

The five metadata tools initially display clearly labeled demonstration records. Choosing Fetch connects directly from your browser to Wikimedia and replaces those records with a live response. The text extractor likewise imports only when you choose its import action. This site does not silently send an article query on page load.

Requests are bounded. Some tools expose another batch when the source reports continuation; another batch replaces the currently displayed rows rather than claiming a complete crawl. Rate limits, missing articles and unavailable network responses are reported rather than being disguised as successful empty results.

For the underlying definitions, see the official MediaWiki Action API documentation and Wikimedia pageview API reference.

Find your calculator

Start typing to explore the library.