Openstreetmap Hotel Poi Extract Data: Dataset Structure and Field Coverage

Every row is one mapped property: a point node, or a building footprint stored as a way or relation, carrying whatever accommodation attributes contributors recorded - name, stars, rooms, beds, address, phone and website among them.

The scale is now genuinely useful for analysis. Taginfo counted [452,341 objects tagged tourism=hotel]() as of 2026-08-21 - 255,658 nodes, 190,479 ways and 6,204 relations - alongside 329 tourism=resort objects and hundreds of thousands of hostel, guest_house and apartment entries. That makes the OpenStreetMap Hotels via Overpass Turbo route the largest free geocoded hotel inventory in existence, at a quality score of 8 in Datadory's catalog.

What the table does not carry matters too: no room rates, no occupancy, no brand affiliation codes. It answers where supply sits, and leaves demand and pricing to the statistical sources covered further down.

The whole workflow runs in a browser tab against the free Overpass API - no account, no key, no quota beyond fair-use etiquette.

1. Open Overpass Turbo at - the interactive query IDE that Datadory lists as this dataset's access route.
2. Scope the geography. Zoom to your study area and use the {{bbox}} shortcut for a city or region, or an area filter for a whole country.
3. Filter the accommodation tags. For a hotel-only census, match tourism=hotel on nodes, ways and relations; broaden the regex to hotel|resort|hostel|guest_house|apartment when the competitive set includes non-hotel stays. The headline figure of 452,341 objects refers strictly to the tourism=hotel value.
4. Collapse footprints to points with out center, so every property - including building-outline ways - lands as one latitude/longitude row.
5. Run and eyeball the result on the map pane before exporting; obvious mis-tags surface immediately.
6. Export in the format your stack wants: GeoJSON for GIS, CSV for spreadsheets and BI loads, or the raw JSON and XML responses for pipelines - all four are supported outputs.
7. Respect the shared servers. Public Overpass instances ask light users to stay under roughly 10,000 queries and 1 GB downloaded per day; a national hotel extract fits comfortably, a continental one does not.

A minimal hotel-only query looks like this:

[out:json][timeout:60];
(
  node["tourism"="hotel"]({{bbox}});
  way["tourism"="hotel"]({{bbox}});
  relation["tourism"="hotel"]({{bbox}});
);
out center;

tourism=hotel dominates the accommodation namespace with 452,341 objects, while tourism=resort appears on only 329. Resort tagging is simply rare in practice, so treat it as a supplement for flagging large leisure complexes rather than a substitute for the hotel value. Hostel, guest_house and apartment values add hundreds of thousands of further properties when your market definition includes budget and short-stay supply.

The OSM value hierarchy maps onto the official classification used by Eurostat - Capacity of Tourist Accommodation (tour_cap_nat), which splits European supply into three NACE classes. Aligning your Overpass regex with those classes is what makes an OSM extract comparable to official supply statistics:

Note what the headline number excludes: campsites, motels and aparthotels sit outside tourism=hotel, so a total-market study needs the wider regex rather than the single tag.

You should, because the official benchmarks are free and sit one join away.

Outside Europe, the pattern repeats with different sources. The US NTTO I-94 program publishes free monthly arrival workbooks built from ADIS records covering over 355 million travelers, and the World Bank's ST.INT.ARVL series covers roughly 265 economies of inbound arrivals - context layers for weighting any site-selection score.

In practice most teams publish derived products instead of the raw table: density maps, per-region supply counts, distance-to-competitor features inside a model. Aggregated statistics and visualisations are the conventional safe harbour; legal review belongs in the process whenever the extract itself would ship to clients.

To put the supply picture into motion, pair your extract with the Eurostat demand tables: tour_cap_nat fixes each country's bed-place denominator, and tour_occ_arn2 supplies the regional nights and arrivals trend that turns a static POI map into a market indicator.