The problem starts at the source. The booking platforms we draw from file almost everything under a single generic label — “accommodation” — with no field for property type and often no guest count. A tiny house, a 12-person villa and a city apartment can all arrive looking the same. So the sorting is on us. This is what has survived that sorting so far:
Step one: the name-and-type check
We read every listing’s name and type and classify it into a category: tiny house, cabin, pod, shepherd hut, yurt, dome, tree house, A-frame, trullo and a few more (the full family, with a live breakdown, is in what counts as a tiny house). Names are reliable here because hosts advertise the format proudly — “Tiny House”, “Shepherd’s Hut”, “Glamping Pod”. Anything that reads like a hotel, a 4-bedroom house or an apartment block is rejected.
Step two: the size gate
We pull the guest count and bedroom count where the platform provides them — and where it doesn’t, we read them out of the listing name (“sleeps 8”, “3BR”). Anything that sleeps more than six or has four-plus bedrooms is not a tiny house, whatever it calls itself, and is dropped.
Step three: human verification
For a growing share of listings, a researcher checks the actual property — opening the live listing or other sources to confirm it really is a small, standalone stay, and recording the source. Those carry a short “why it’s on the atlas” note with a link. When we can’t confirm something, it stays off the map rather than being guessed onto it. We would genuinely rather list fewer houses than list a hotel room with “tiny” in its name.
What that means for the numbers you see
Every figure on this site — the stay counts on destination pages, the medians in our best-of lists, the tables in the journal — is computed live from this verified pool, never typed into an article. Prices are sampled totals for near-term dates, shown as medians and quartiles rather than “from” teaser rates; guest ratings come from the booking platforms and we average only stays that actually have them. When the pool changes, every page changes with it.
Where it can still go wrong
It isn’t infallible. Hosts rename properties, platforms shuffle their data under us, and a name check can misread a creative title. Two honest examples of our limits: feature tags (hot tub, pets, off-grid) only exist where hosts state them explicitly, so they undercount reality; and regions whose local rental market never reaches the global platforms — Denmark is one — are underrepresented no matter how good they are. The whole point of the atlas is the curation, so if you ever spot something that doesn’t belong, tell us and we’ll recheck it.