How to follow a country's news without speaking the language

The problem with the usual sources

When a desk needs to follow a country nobody on staff can read, the standard toolkit has three layers, and each one filters out most of what you want. International wires carry a country's news only once it matters internationally, which is late by definition. English-language editions (the Japan Timeses and Deutsche Welles of the world) are written for you, which means they are a selection, often hours behind their domestic siblings. And social media gives you whoever tweets in English, which is a persona, not a press corps.

Meanwhile the country's own newsrooms are doing the work in real time: deciding, minute by minute, what is worth interrupting their readers for. That decision layer is the thing you want. The only problem is the language it arrives in.

The signal: their pushes, not your searches

A push notification is a newsroom's own judgment, compressed to one sentence and timestamped. Watch every notification a country's apps send and you are watching its domestic news agenda form, before any of it is translated, picked up by a wire, or explained to outsiders. In our twelve-month dataset, 77.6% of 537,376 notifications were not in English, and the stories that eventually crossed three or more countries (6,566 of them) almost all broke in someone else's language first.

AO.news makes that stream readable: every push is machine-translated to English within seconds of being sent, with the original text preserved, and notifications about the same story are clustered so you can see how many independent outlets are on it. That last number, consensus, is what replaces your missing language instinct. You cannot judge a Hungarian headline's register, but you can see that seven Hungarian newsrooms pushed the same story inside twenty minutes.

The workflow, step by step

1. Build the country column. In the deck, create a column and select the country's sources with one click. Set consensus to 2: you now see every story at least two of its newsrooms considered push-worthy, in English, and nothing else.
2. Build the "everything" column next to it. Same sources, consensus off. This is the raw domestic agenda: the weather warnings, the celebrity gossip, the football. You will not read it constantly, but it is where you go to understand what the country is actually talking about.
3. Add a beat filter if you have a beat. A word filter like (nuclear OR reactor) AND !sport runs against the English translation, so one query covers the language you cannot read.
4. Route the corroborated column into Slack. With the Slack integration, the 2+ column becomes a channel the whole desk sees, each alert carrying the source, the translation and a permalink.
5. Open originals when it matters. Clicking a notification opens the article, with Google Translate applied if you want it. The translation gets you to "this is a story"; the original, plus a phone call, gets you to publishable.

Worked example: Hungary, without Hungarian

Take a real case from this year's data. On 12 April 2026, Hungary voted in the election that ended Viktor Orbán's 16 years in power. It became one of the year's biggest stories, pushed by 110 apps in 22 countries. Here is what a desk with a Hungary column saw, reading only English:

• The morning previews came from everywhere: Italian, Finnish and Austrian apps pushed "decisive day for Orbán" pieces from dawn. That told you the stakes, not the story.
• The story itself came from the Hungarian apps. Telex pushed the candidates casting their votes, Magyar's pledge to cap his own time in office at two terms, and Orbán declining to message his challenger, all within the morning, all translated within seconds of each send. Népszava and 24.hu filled in the record and the polling-station detail. This is the granularity no wire carries.
• The consensus number told the arc. Through the day the cluster climbed toward 110 apps as French, German, Nordic and Brazilian outlets joined on the result. Watching that count rise is watching a domestic story become a world story, in real time.
• The permalinks settled "who said what, when". Every notification in the cluster carries its send time and source, which is exactly the audit trail you need when you write "Hungarian media reported at 19:40 that...".

None of this required a stringer in Budapest. It required the judgment of Budapest's own newsrooms, translated and counted.

Verifying what you cannot read

Machine translation is good enough to know a story exists and roughly what it claims. It is not good enough to quote from, and it cannot tell you an outlet's politics. Three rules keep a language-blind desk safe:

• One push is a claim; three are a story. Never move on a single translated notification. Consensus is the filter that needs no cultural context.
• Learn the country's source map once. Ten minutes with a colleague or a country handbook: which outlet is the public broadcaster, which is the tabloid, which is government-aligned. The deck shows you every source's name and icon on every alert, so the map stays usable.
• Confirm before quoting. Use the translated push to find the story, the original article for the facts, and a human, your own reporting, or the outlet's English service for anything you will attribute in print.

Honest limits

Coverage is uneven: our deck is deepest in Europe and the US, thinner elsewhere, and a country's press freedom sets a ceiling on what its newsrooms will push at all. Machine translation flattens nuance and occasionally mangles names. And pushes skew to breaking news, so slow investigative work surfaces only when it makes the domestic front pages. For the countries we cover well, the per-country pages show exactly what the stream looks like: Sweden, Norway, Denmark, Finland, Germany, Switzerland, France, Belgium, Italy, Spain, United Kingdom, United States, Israel, India, Brazil, New Zealand.