Skip to content
How Outrouted works

What Outrouted's trip data comes from

Every map and every number on this site comes from somewhere. This page names the sources category by category, so you can judge how grounded an answer is instead of taking it on trust.

Updated 24 Aug 2026about 4 minutes

1.Maps and terrain

Three sets of tiles, for three jobs. All of them are open and keyless, which is the reason they can be used on a site that runs no third-party scripts at all.

WhatSource
The map inside the productOpenStreetMap data, served as raster tiles by CARTO
The atlas on the marketing pagesOpenStreetMap data, served by OpenFreeMap
Relief, hillshading and 3D terrainOpen elevation tiles hosted on Amazon's open data registry

2.Places, cities and coordinates

Curated destinations, their highlights and their hotels are Outrouted's own catalogue. Countries, cities and airports outside that curated set are generated from open geographic data, which is what makes anywhere in the world plannable rather than only the places somebody has written up.

Turning typed text into coordinates goes through LocationIQ, Geoapify and geocode.maps.co, tried in order until one answers. A highlight's position is hand-checked where it matters and geocoded otherwise, and a geocoded hit is only kept if it lands inside the destination's own country and within ninety miles of it. A place that survives neither check has no position, and Outrouted draws nothing rather than a pin in the wrong country.

3.Accommodation

Live stay prices, and the whole booking flow if you book a room here, come from Nuitee's liteAPI. Where a live quote is not available for those dates and that place, the catalogue's own nightly rate is used and the figure is an estimate.

4.Flights

Nuitee first, with SerpApi's Google Flights results behind it. Both can only answer a dated, airport-to-airport question, so anything vaguer than that is estimated from distance instead.

Flights are priced here rather than sold.

5.Exchange rates

European Central Bank reference rates, fetched through Frankfurter. When that is unreachable a pinned table is used instead, and it is indicative rather than a rate anybody trades at.

6.Weather

Seasonal norms held per destination: typical highs and lows, wet days, daylight hours. Not a forecast, and deliberately not one. Somebody planning six months out wants to know whether to bring a coat, and a ten-day forecast cannot tell them.

7.Photography

Destination photographs come from Pexels. The photographs on the landing page's own map are held in this repository rather than hot-linked, so the first screen anybody sees does not depend on somebody else being up. Their credits are on the about page.

8.The language model

A hosted large language model chooses which places belong in a trip, in what order, and how many nights each gets. It is sent the trip's inputs and the pool of candidate places, and no account identifiers. It never produces a number that appears on the screen.

9.When a provider changes

Every provider above sits behind an interface rather than being wired into the product, so swapping one is a configuration change rather than a rewrite. The health endpoint reports which arrangement is actually running, and this page is updated when one changes.

None of them is told which page you are on, and none of them is paid for placement in a result.