Indoor Photo Geolocation: Why It Is Harder

Almost everything written about reading a location from a photograph assumes you are outside. Signage, architecture, vegetation, road markings, the angle of the sun — the standard toolkit is an outdoor toolkit. Step through a door and nearly all of it disappears at once.

That is why interiors are the reliable way to defeat this kind of guessing, and it is also why they are more interesting than they first appear. A room is not empty of geography. It just holds a much smaller and stranger set of clues, most of which people never consciously notice because they have only ever lived among one version of them.

Consider what an outdoor frame gives away without trying. The sky reports weather and, through shadow length and direction, something about latitude and season. Plants report climate. Signs report language and often the exact administrative region. Vehicles and road markings report a driving side and a set of national conventions. Buildings report a construction era and a regional style. A typical street scene is dense with independent evidence, which is why estimates from it are frequently good — the reasoning is laid out in how AI guesses where a photo was taken.

A bedroom offers a white wall, a bed and some soft furnishings. None of those vary reliably by country. The frame is not low quality, and the model is not underperforming; the information simply is not present. This is the same reason certain outdoor shots also fail, a pattern explored in what makes a photo hard to geolocate.

• —Sockets and plugs. The strongest single clue in most rooms. Roughly fifteen plug and socket types are in general domestic use, and each maps to a defined set of countries.
• —Light switches. Rocker versus toggle, plate size, and mounting height all vary regionally, and switches appear in an enormous share of interior photographs.
• —Heating. Tall panel radiators under windows read as northern European; a wall-mounted split air-conditioning unit points somewhere warm; a visible boiler or a radiator's absence entirely is itself a signal.
• —Paper and books. A sheet of A4 is noticeably narrower and longer than US Letter. Once you know the proportion, a page on a desk is a continental-scale hint.
• —Doors and windows. Inward-opening tilt-and-turn windows are a strong central and northern European marker; sash windows point elsewhere entirely.
• —Text on anything. A book spine, a cereal box, a remote control, a plug socket label. Language plus script narrows the field faster than any other observation.

The socket deserves its reputation. It is standardised at national level, can stay unchanged for 50 years or more, appears at a consistent height and is rarely obscured by furniture. Its shape correlates strongly with mains voltage regions, which split broadly into 100–127 V and 220–240 V families. One clearly visible outlet regularly does more work than the rest of the room combined.

Paper size is the clue people find most surprising. Almost the entire world uses the ISO A series; North America uses Letter, which is wider and shorter. The proportional difference is subtle in isolation and obvious once you have compared them, and it separates two continents from a document lying on a table.

There is a real irony here. The rooms most likely to be photographed by travellers — the ones where you might most want to know the country — are the ones stripped most thoroughly of national identity. That is the product working as intended. International consistency is the promise a chain is selling, and the visual result is an interior with almost no fingerprint. If you want to test a tool's honesty, a hotel room is the ideal input: a system that names a confident city from one is telling you something about itself rather than about the photograph.

The practical lesson for anyone uploading their own pictures is simple. If an interior matters to you, include a window, a doorway, a socket or anything carrying text. That advice generalises, and the fuller version is in best and worst photos to upload. The way scene type shifts the whole problem is covered in urban vs rural photo geolocation.

Interiors are the honest test of any geolocation tool, because they remove the easy wins and leave only the reasoning. A model that admits a plain room could be almost anywhere is behaving correctly. One that produces a street address from a beige wall is not being clever — it is guessing and hiding it.