Serbian news API

What the coverage looks like

Unusual among major languages in being written in two scripts, often by the same publisher, sometimes within one site. This is the fact most likely to silently halve your results.

Outlet spread is the number worth watching here, not volume. 51 distinct publishers at roughly 22 articles each describes a very different information environment than the same total produced by a handful of prolific sources — and it is the difference between reaching independent reporting and reading one newsroom repeatedly.

In this sample the most active publisher is naslovi.net, carrying about 23% of Serbian articles on its own, and the three busiest — naslovi.net, b92.net, blic.rs — account for roughly 43% between them. That is a concentrated distribution, so a query returning a lot of Serbian results may be reaching fewer independent newsrooms than the count implies. Scope by outlet if you need breadth rather than volume.

What is the ISO 639-3 code for Serbian?

srp. Pass it as the language filter on any endpoint. Serbian is written in Cyrillic and Latin.

What does Cyrillic and Latin script mean for Serbian search results?

Cyrillic and Latin are both standard and fully interchangeable, and a keyword query in one script matches nothing written in the other. Mutual intelligibility with Croatian and Bosnian also means related coverage sits in adjacent language facets.

Which retrieval mode should I use for Serbian news?

Never rely on a single-script keyword query. Semantic mode handles the split; if you must use keyword mode, query both scripts and merge. Consider querying Croatian alongside Serbian when your interest is regional rather than national.

Bound every request with a time range regardless of mode. Meaning-based retrieval always returns its best guesses, so an unbounded query produces a confident-looking result set anchored to nothing in particular. If results feed a model rather than a person, collapse to stories as well — one event covered by thirty outlets should cost you one item, not thirty.

Recent headlines in Српски / Srpski

Pulled from the index on 2026-07-31, shown in their original script so you can see what the API actually returns rather than a translated approximation of it.

• Мацут: Врањска Бања има велики потенцијал, циљ је да додатно ојачамо бањски туризам politika.rs
• Појачан саобраћај на више граничних прелаза у БиХ politika.rs
• У Србији сутра веома топло, температура до 40 степени politika.rs
• Mrtav muškarac zatečen na kolovozu Užas u Zemunu: Imao povrede glave i noge, policija na licu mesta naslovi.net
• Блокадери упутили иронично „извињење" просветарима politika.rs
• Nova "Blic žena" donosi inspirativne priče, beauty savete i još mnogo zanimljivog sadržaja zena.blic.rs

Outlets publishing in Serbian

The most active publishers in this language, by article count. Use these as an outlet filter when you want to pin a particular market or editorial position rather than taking the whole language.

naslovi.net · b92.net · blic.rs · politika.rs · informer.rs · telegraf.rs · danas.rs · tanjug.rs · novosti.rs · alo.rs · rts.rs · rtrs.tv

What catches people out when querying Serbian?

The two-script problem is the one that silently halves result sets. Cyrillic and Latin are both fully standard, sometimes on the same site, and a keyword query in one matches nothing written in the other. A team can run this for months believing coverage is thin when they are seeing roughly half of it.

A worked case

Never rely on a single-script keyword query. Semantic mode handles the split; if keyword mode is unavoidable, query both scripts and merge the results. When your interest is regional rather than national, query Croatian alongside Serbian, since mutually intelligible coverage is split across separate language facets.

Where this language leads

Strong on regional politics, and on Balkan affairs generally where coverage in any other language is thin and usually second-hand.

Who queries it, and why

Queried for Balkan regional politics, where coverage in other languages is thin and usually second-hand. The dual-script issue means teams routinely see half of what exists and conclude the coverage is sparse.

Reading these numbers

Provisional. Measured from one ingestion adapter's public feed rather than from the index itself. Outlets and headlines are real; the counts are the right order of magnitude and are not coverage guarantees.