About the data and how the atlas was made
The atlas is built on open databases put together by linguists from all over the world. Here is where everything comes from, what is certain and what is only an estimate.
What is on the globe
Each dot is one language. The list of languages comes from Glottolog, an open database of the world's languages run by the Max Planck Institute for Evolutionary Anthropology in Leipzig. The atlas takes everything Glottolog lists as a language and gives a location: 7,967 languages.
A dot stands where Glottolog places the language, usually in the middle of the area where it is spoken. Even a language spoken all over the world has just one dot, at its home.
Left out are only Glottolog's special bins: bookkeeping entries, languages known only by name and languages we know so little about that they cannot be classified. Esperanto, Interlingua and Interslavic have no dot because they did not arise in any one place; they are still in the list and in search.
Language or dialect?
There is no sharp line between a language and a dialect, and history and politics often play a part. The atlas follows Glottolog's decisions. That is why Bavarian has a dot but Australian English does not – Glottolog counts it as a variety of English. Dialects are listed on the language card, and searching for one finds its language.
Serbian, Croatian, Bosnian and Montenegrin are very easy to understand for each other's speakers and rest on the same dialect base, Shtokavian. Linguistically they are close standard varieties of one shared language; officially and socially they are separate languages today. Glottolog lists them as one language, Serbo-Croatian, which has a dot on the globe. The atlas also has Serbian and Croatian as separate languages with a greeting, but they have no dot of their own. Hmong, on the other hand, is several languages in Glottolog and one entry in the atlas. That is why the list has 7,970 entries and the globe 7,967 dots.
164 languages with a detailed card
For 164 selected languages the atlas gives a greeting, an approximate pronunciation, a note, the number of speakers and the area of distribution. These are compiled by hand and the notes are fact-checked.
The pronunciation guide is only approximate. Speaker numbers are rounded estimates that include second-language speakers.
The orange area on the globe is drawn by hand and approximate, not measured. There are no open data with exact language boundaries for the whole world.
The Play pronunciation button uses the device's speech synthesis; for some languages the device has no voice.
The illustrations show the typical landscape and vernacular architecture of the region the language comes from. They do not depict particular places.
Details for the other languages
Vitality: the UNESCO scale (safe to extinct) based on Glottolog data.
Typology (word order, cases, tone, consonant inventory… – 24 features): WALS Online, the World Atlas of Language Structures (Dryer & Haspelmath 2013). WALS describes a sample of about 2,600 languages; the others have no value. For the map some values are merged into groups; the language card always shows the exact WALS value.
Number of sounds: PHOIBLE. Sample text: Article 1 of the Universal Declaration of Human Rights (UDHR in Unicode).
Speaker numbers: Wikidata; where missing, an estimate from Unicode CLDR.
Script and official status: Unicode CLDR (languageData, territoryInfo). Official status is national, de facto or regional; CLDR records it only for languages whose country data it tracks, so it is often missing for minority languages even where they have some official standing. For other languages the script map (Maps › Script) uses the script estimate from CLDR likelySubtags, so that the map shows a script for about 6,800 languages in all. The language card marks an estimate as such.
Relatives and the family tree: the Glottolog tree, with Glottolog's branch names.
Nothing is made up: if a source has no data, the card does not show it.
Languages of antiquity and Language stories
The Atlas writes the pages on ancient civilizations and the Language stories from the scholarly literature listed at the end of each page, and checks the facts before publication. The photographs come from Wikimedia Commons and have free licences (CC BY, CC BY-SA, CC0 or public domain); the author and licence are given with each image. Cuneiform signs and Anatolian hieroglyphs are rendered with the Noto fonts (SIL Open Font License).
Questions for the AI
If the search does not understand a question, you can ask an AI (Claude Sonnet 5.5 by Anthropic). The question text is sent via the atlas server (Netlify) to Anthropic; the atlas does not store questions. The AI looks in the atlas data first (Glottolog, Unicode CLDR, atlas texts); what the atlas does not have, it looks up on the web (Anthropic web search) and lists the web sources under the answer. It may still be wrong, so every answer carries a note. It answers only questions about languages, scripts and language families.
Map
Country borders and land come from Natural Earth (simplified, scale 1:110 million). Relief and the sea floor come from Natural Earth and NOAA ETOPO1.
Data versions and licences
| Glottolog – languages, families, vitality, dialects | downloaded 22 September 2026 | CC BY 4.0 |
| WALS Online (2013) – grammar | downloaded 1 October 2026 | CC BY 4.0 |
| PHOIBLE 2.0 – sounds | downloaded 1 October 2026 | CC BY-SA 3.0 |
| Wikidata – speaker numbers | downloaded 1 October 2026 | CC0 |
| Unicode CLDR 48 – language names, script, official status, estimates | Unicode License | |
| UDHR in Unicode – sample texts | udhr package 6.0.0 | MIT |
| Wikimedia Commons – photographs for the civilization pages | with each image | CC BY, CC BY-SA, CC0 |
| Natural Earth, NOAA ETOPO1 – map and relief | public domain |