Methodology

What we do and do not do

Realidad Multipolar does not promise to remove bias: no system, human or automated, can. We offer a map of perspectives: what each geopolitical prism says about the same event, where they agree, where they clash and what is missing.

Our commitment is that the method is visible. You may disagree with our choices, but you can see what they are and why we made them.

How we classify sources

We classify each source by its origin, never by whether it "tells the truth". A prism says where a view comes from, not whether a source is reliable or what everyone there thinks.

Atlantic: the United States, the European Union, the United Kingdom and NATO. Russia: state and non-state voices from the Russian Federation, including independent Russian outlets working in exile. China: the People's Republic of China, including Hong Kong outlets. Global South: India, the Arab world and the Gulf, Africa and Latin America. Cross-cutting: wire agencies and multilateral bodies.

We also state each source's type (wire agency, public or state media, private media, independent media, think tank or official source), who owns it, how it is funded and how far we have verified this. It is all in the sources page.

Editorial rules

Three prisms. A topic is presented as compared when at least three prisms cover it; when fewer do, we say so.

Every claim has its source. Everything we attribute to a prism links to the original article or statement. Quotes are short and in their original language, with our translation when needed.

We summarise and link; we never copy. For the full content we always send you to the original outlet.

Agreements and clashes. A shared fact must be backed by sources from at least two prisms. A clash compares what sources say about the same fact at the same moment; when they describe different moments of a developing event, we say so. When a primary record settles a factual clash, the editor marks it as a verified fact, with a link to that record; the automatic drafting never does.

Sanctions. We do not read outlets we know to be under current European Union sanctions, and for those prisms we prefer official sources. The source-by-source check against the official lists is pending.

Diversity within a prism. Where possible we include critical voices within the same prism. No country speaks with one voice.

What a prism without coverage means

When a prism shows no coverage, it means we found no coverage from that prism among the sources we consulted for that edition. It does not mean nobody from that prism published anything: it can depend on which sources are in our catalogue, a source that failed that day, the time we checked or how we grouped the articles.

Locator maps

When a topic happens mainly in one place, the day's list places it on a small map. The maps are our own drawings, made from Natural Earth's public-domain data, with no flags or prism colours; each topic has at most one place; when it has none, the map shows its region, or the world for a global topic.

Disputed territories (Crimea, Western Sahara, Kosovo, Palestine —the West Bank and Gaza—, Taiwan, Northern Cyprus, Somaliland and Kashmir) are always drawn with neutral hatching, assigned to no country and never highlighted. A topic that happens mainly in one of them shows it hatched, with its name. A map only locates the event and implies no position on any territory's sovereignty.

How we use artificial intelligence

AI is a supporting tool, not the editor. It groups the articles about the same event, drafts structured summaries by prism, translates into English, French, Russian and Chinese and checks every attribution and quote again against the text we stored from the original, flagging doubtful ones. When we could not store the page, the check is made against the summary the outlet publishes in its news feed; for long texts only the beginning is checked. Gathering the articles and ranking the topic proposals are done by fixed rules, without AI.

The editors review the topics they choose. In addition, every day the system picks some events covered by at least three prisms on its own and drafts them; the editor is notified and, if the draft passes every check without flags, it is published two hours later unless the editor discards or edits it. Drafts about the arrest, charges or trial of a named person always wait for the editor, and so do drafts that give casualty figures without saying who gives them or hedging them. When the figures are attributed ("according to the police") or hedged ("at least"), the topic may publish on its own and says they are provisional. Those topics carry the label "Published automatically, without human review". When we can, we keep a copy of each source's text from the moment we read it: if the outlet changes the article later, the check was made against the version we read.

Errors and corrections

If you spot a wrong attribution or a classification you think is mistaken, write to hola@realidadmultipolar.com. A corrected topic shows the date of the correction. We do not yet publish a list of corrections or the previous text.

Known limits

Choosing the sources is already an editorial decision, which is why we publish it. Prisms simplify complex realities: they are a tool to find your bearings, not a truth. The catalogue does not yet cover every prism equally: today the Atlantic prism is read through European sources, with no source from the United States; the Cross-cutting prism, only through UN News, with no wire agencies; and the China prism includes no independent media. AI can make mistakes, which is why we link every claim, check every quote against the text we were able to store and label the topics published without human review.

In Spain, some operators block shared Cloudflare addresses by court order during football matches, so this site may not load then. It does not depend on us. The weekly newsletter still arrives by email.

Last updated: 17 November 2026