What this is
The Albania Atlas is an English-language reference to Albania: its places, its history, its culture, its economy and the practicalities of visiting or living there. It is built as a connected graph rather than a pile of articles — a beach links to its nearest town, that town to its county, the county to the industries that employ it, and those to the port that ships their output. You can arrive anywhere and follow the thread.
It is not a travel blog, and it is not trying to sell you a holiday. Roughly a third of it is about subjects no holiday needs: administrative divisions, the electricity system, the 1997 collapse of the pyramid schemes, the Kanun, the fason textile model, and how to read a population figure that is disputed.
How it is built
Three layers, deliberately separated so that one cannot quietly corrupt another.
- The curated layer. Every entity — 12 counties, 75 settlements, 26 beaches, 28 historic and archaeological sites, 29 natural features, 30 guides, 15 explainers and 12 comparisons — starts as a hand-written record: what it is, why it matters, a plain-language explanation, and what it connects to.
- The measured layer. No number on this site is typed by hand into prose. Population, area and elevation come from Wikidata claims and are rendered with their observation year. Macroeconomic and demographic series come from the World Bank API. Monthly climate normals are computed here from thirty years of ERA5 reanalysis, so every place is measured on the same basis. County boundaries come from geoBoundaries.
- The research layer. For every subject with a Wikipedia article, the full article is fetched and read by a language model that is instructed to write structured sections from that document only. The article is named, linked, and the number of characters read is printed at the foot of the page. It is AI-assisted research over a named source, and it is labelled as such wherever it appears.
The quality gate
Programmatic reference sites fail in a predictable way: they generate a page for every
place name and end up with thousands of entries containing a name and one number. This Atlas
scores every page on what it actually carries — explanation, measured facts, photographs,
research, relationships, climate — and a page that scores below the bar stays in the database
and stays reachable, but is served noindex and marked as a stub. It earns
indexing by becoming useful, not by existing. 167 pages currently carry a deep
research section, and 885 correctly-licensed images are published.
What it will not do
- It will not give legal, tax, immigration, financial, investment or medical advice. Several sections — retirement, property, residence, healthcare — touch subjects where a confident answer from a website is worse than no answer.
- It will not print a price table without a date and a source, because an undated price is wrong within a year and the reader cannot tell.
- It will not call anything the best beach in Europe. Where a ranking is quoted, the publisher and the year are named so you can judge it.
- It will not publish an image whose licence it cannot read. See the photo policy.
- It will not republish other people's journalism. News items show a headline, publisher, date and short summary, and link out.
Who writes it
The Atlas is compiled and edited with AI assistance over named sources, under the rules set out in the editorial policy and methodology. It is not written by Albanian institutions and is not affiliated with any government, tourism authority or business. Corrections are welcome and are checked against the source record — there is a form at the foot of every page.
Reusability
Nothing in the underlying software is specific to Albania. The entity kinds, the relation engine, the quality gate, the climate pipeline and the licensing gate would run a Greece Atlas or a Croatia Atlas unchanged; only the seed data differs. That was a design constraint from the first line of the schema.