Put a missing building on OpenStreetMap with Rapid’s AI suggestions

A magenta outline in Rapid is a building suggestion awaiting a mapper’s decision. Accepting it gives you editable geometry; uploading your finished edit adds it to OpenStreetMap. That gap between suggestion and contribution is where local knowledge becomes useful.[1]

Start with one clearly visible, detached building in an area you know. This walkthrough follows published Rapid and OpenStreetMap documentation; no map edit was uploaded in preparing it. Use a desktop or laptop so you can inspect the outline and its surroundings comfortably.

1. Find a suitable suggestion

Open Rapid and sign in with your OpenStreetMap account, authorizing the editor when prompted. If you need an account, create one on OpenStreetMap’s main website; map editing uses a different account from the OSM wiki.[2]

Search for your neighbourhood and zoom in until individual building corners are easy to distinguish. New mappers should complete the editor’s walkthrough, especially its Buildings and Rapid sections. The help panel opens with H; the keyboard-shortcut reference opens with ?.[3]

Open Rapid Assist and choose an available AI building-footprint dataset, such as Microsoft’s. Inspect a magenta candidate against both the imagery and existing OSM features. Choose a building whose outline is missing. Skip a candidate that duplicates existing mapping, spans several terraced houses or cannot be confidently identified. The OSM guidance specifically flags terrace separation and alignment as weaknesses to examine.[1]

2. Turn the suggestion into a good outline

Select the candidate and click Use This Feature, or press A while the suggestion is selected. It becomes editable. Features you have not accepted are excluded from the upload. If the candidate is unsuitable, Remove This Feature, or D while it is still a suggestion, dismisses it.[6]

Check the whole polygon before tidying individual corners. Select the outline and use M to move it or R to rotate it. Q squares its corners; apply that only when the real building has right angles. Use Undo if a correction makes the outline worse. These operations depend on what is selected, so keep the building selected rather than the map background.[3]

The position to capture is the building’s footprint. In an angled aerial photograph, the roof can appear displaced from the walls at ground level. A neat polygon tracing that roof can therefore be misplaced. Where the base is visible, use its corners to position the outline, accounting for perspective. If the correct location cannot be resolved from imagery and local knowledge, choose an easier building for this first contribution.[4]

3. Give it the right meaning

Inspect the building’s tags in the feature panel. For a confirmed building whose architectural type you do not know, building=yes is the useful general classification. Use a more specific type only when you can establish it. A footprint alone does not establish that the premises contain a shop, café or clinic.[5]

Building type and present use are separate facts in OSM. An old hospital converted to another purpose can retain building=hospital, while amenity=hospital describes an active hospital. The same distinction explains why a retail-shaped building is insufficient evidence for an operating supermarket. For this exercise, a sound outline with a general building tag is a complete contribution; an invented name or business category adds an error.[5]

4. Upload and find your contribution

Review the validation panel and resolve issues affecting your edit. Zoom out once to check how the building relates to neighbouring roads and structures, then press Save. Read the proposed changes and enter a concise changeset comment describing what you did. For example, if accurate: Added one missing building outline; checked alignment against aerial imagery and local knowledge. Complete the upload and wait for confirmation.[6]

On OpenStreetMap, open History for the area and find the changeset with your account name and comment. Inspect its added feature. The changeset page gives the contribution a persistent record and lets other mappers understand what changed. Keep its link if you want to return to the work.[7]

If the rendered map still looks unchanged, check that record before adding the building again. Rendered map tiles can lag behind uploaded data. A successfully uploaded feature does not need a duplicate while the picture catches up.[8][7]

If no suitable suggestions appear, inspect which datasets cover your area and leave uncertain candidates alone. The aim of this first session is one building whose existence, position and tags you can explain. The next time you pass it, you will know exactly which small part of the shared map came from your judgement.

Report an error or suggest a correction