OpenStreetMap has mapped the world without a central schema, a fixed list of feature types, or a committee approving each new kind of thing, and the mechanism that made that possible fits in five characters: key=value. A tag is a pair of text strings, a key naming the topic and a value carrying the detail, attached to a mapped element. highway=residential says this line is a road of the residential kind. amenity=cafe says this place serves coffee. From those two strings, multiplied by hundreds of millions, the entire map is described: its roads and rivers, its opening hours and speed limits, its benches, borders, and bike lanes. This guide explains the tag system as a reader and a future editor should understand it: what elements tags attach to, how keys and values divide the descriptive labor, the few house rules that keep the system coherent across millions of volunteers, the colon namespaces that organize big key families, and the two institutions, the wiki and Taginfo, where the conventions are written down and counted. It builds on our open map database explainer, which covers the project as a whole; this article is about its grammar.
Home / GIS and Open Map Data for Beginners
OpenStreetMap Tags Explained: How key=value Describes the Whole World

Photo: Kha Ruxury / Pexels
Key Equals Value: The OSM Tags Checklist
Check off each step as you finish it. Your progress stays on this page.
0 of 7 complete
0% done
The Atom: A Tag on an Element
Tags never float free; they describe one of the map's three element types. A node is a single point, a tree, a bench, a cafe's location. A way is an ordered line of nodes, a road, a river, a building's outline. A relation is a grouping of elements with roles, a bus route, a boundary, a long-distance trail. Each element carries a set of tags, and the set is the description: a building outline might carry building=yes, name= a person's name, and an address family, and together those tags are everything the map knows about that building. Two consequences shape the whole system. First, an element's primary tag is what makes it the thing it is: strip highway=residential from a way and it becomes an anonymous line, which is why editors speak of tagging a feature rather than drawing one. Second, description is additive and open-ended: anyone can add a relevant tag to any element, a wheelchair access note to the cafe, a surface tag to the road, without asking the schema's permission, because there is no schema to ask. The map's richness and its occasional messiness are both born in that freedom, and the house rules below are how the project keeps the freedom productive.
Keys Name the Topic, Values Carry the Detail
The equals sign divides labor. The key names the subject: highway, amenity, name, maxspeed, surface. The value gives the specific: residential, cafe, the actual name, the actual limit, asphalt. Values come in a few species worth recognizing. Some are enumerations, drawn from a shared vocabulary the community maintains: the values of highway, from motorway down to path, are a ladder every editor and every map style learns. Some are free text, above all name, where the value is whatever the sign on the ground says. Some are numbers, ideally with units made explicit where the bare number would be ambiguous. And some are effectively booleans, yes and no, answering whether a property holds: dogs allowed, wheelchair accessible, covered. A well-tagged feature usually carries one primary tag that says what it is and a supporting cast that says what it is like: the cafe gains cuisine=, opening_hours=, outdoor_seating=yes, each a key doing its own topic. Reading a feature's tags is therefore like reading its dossier, line by line, each line a small fact with its topic labeled. The map style you look at is just a rendering of dossiers like these, drawn by rules that say a way with highway=primary gets the orange line.

The House Rules: One Key Once, Semicolons for More
A system with no schema survives on conventions, and three of them do most of the work. First, an element carries each key at most once. There is no second name tag, no repeated highway key; the one instance must say the thing properly, which forces clarity about which fact is the fact. Second, where reality genuinely holds several values for one key, a road serving two route numbers, a restaurant cooking two cuisines, the convention is a semicolon-separated list inside the single value, cuisine=italian;pizza, rather than duplicate keys. Third, and most important, do not invent keys lightly. The vocabulary is a commons: a key means what the community uses it to mean, and a private invention describes your feature to nobody, because no map style, router, or search engine will ever read it. The working editor's habit is to check before coining: search the wiki for the concept, see the key the world already uses, and use it even if your own coinage feels cleverer. Consistency across millions of strangers is the entire value of the database; a tag only one person understands is a note in a diary, not a contribution to a map.
Colons: The Namespace Families
As the vocabulary grew, keys acquired structure: the colon separates a family name from a qualifier, building families of related keys under one roof. The address family is the everyday example: addr:street, addr:housenumber, addr:city, addr:postcode, each a facet of one address, grouped by their shared prefix. Language qualifiers work the same way: a name in the local language stands in name, and its versions in other languages ride alongside as name:en, name:fr, as many as editors care to add, which is how one street carries its names in a dozen tongues without conflict. Other families organize by source, by lifecycle, by the side of the road a property applies to, and the pattern generalizes: the part before the colon is the topic, the part after refines it. For the reader of tags, colons are a gift: they announce that a set of keys belongs together, and learning the big families, addr, name with its languages, and the contact and source families, decodes most of what a richly tagged feature is saying. For the writer, they are the sanctioned way to add detail without inventing chaos: refine an existing family's key before founding a new one.
Where Conventions Live: The Wiki and Taginfo
Two institutions keep the vocabulary honest, and every serious user visits both. The OpenStreetMap wiki is the written record of the conventions: a page for nearly every key and common value, saying what it means, what it applies to, how it combines with other tags, and where its usage was debated. New tagging ideas go through a proposal process on the wiki, discussed and voted, though the system's real law is practice: a tag becomes standard by being used, documented, and picked up by the software that renders and routes. Taginfo is the census: it counts actual usage across the live database, how many elements carry each key and each value, where in the world they cluster, and how usage grows. Its scale tells the story of the whole project: the wiki's figures, as of May 2026, describe a vocabulary of over 100 thousand distinct keys and over 190 million distinct tags in use. When you are unsure whether a key is the standard one, Taginfo settles it with numbers: the key used two million times is the convention, and the key used eleven times is someone's experiment. Check the wiki for meaning, Taginfo for prevalence, and your tagging will read like a native's.
That pair of checks takes under a minute, and it is the entire difference between contributing to the map and merely marking it. The wiki tells you what the community means; the census tells you what the community does; where the two agree, which is nearly everywhere that matters, you are standing on the standard, and your edit will be read by every map, router, and search engine that consumes the database, which is to say by the world.
Reading Tags Like a Mapper
Fluency comes from reading real dossiers. A way tagged highway=path, surface=gravel, bicycle=yes is a gravel path bikes may use; the map will draw it thin and the router will send a cyclist down it gladly and a car never. A node tagged amenity=bench is a bench, full stop, one tag doing all the work. A building with building=apartments, building:levels=5, and a full addr family is a five-story apartment block at a known address. A cafe with amenity=cafe, cuisine=coffee_shop, opening_hours=, wheelchair=yes, internet_access=wlan is a small portrait of a business, written in five lines of grammar. Notice how each dossier combines one identity tag with descriptive followers, how the colons cluster the address, how yes values answer access questions a traveler will actually ask. That is the whole art. If you go on to edit, and our guide to adding a place to OpenStreetMap walks through your first contribution, you will write dossiers like these yourself, in the vocabulary the wiki documents and Taginfo counts, adding your lines to the largest description of the world ever assembled one key and value at a time.
Beginner Mistakes the Census Catches
Taginfo's counts quietly document the ways new editors go wrong, and three mistakes account for most of the wreckage. The first is inventing a key for a concept that already has one, describing a drinking fountain with a private coinage when the community's key sits in the wiki waiting; the feature becomes invisible to every renderer and router, mapped and yet unmapped. The second is tagging for the renderer: bending tags until a favorite map style draws the picture you want, misclassifying a path so it shows in a preferred color, which corrupts the data for every other consumer to please one picture. The community's rule is absolute: describe the world as it is, and let the renderers make their own choices. The third is the capitalization and spelling drift, keys written with capitals, spaces, or near-miss spellings that fragment the vocabulary into dozens of one-off variants the census can count and no software will ever read. All three mistakes share a shape: they treat the tag as a label for the editor rather than as a sentence in a shared language. The cure is the habit this article has taught from the start: before you write a tag, read how the world writes it, in the wiki for the meaning and on Taginfo for the numbers, and your lines will join the description instead of whispering beside it.
Related Guides
This guide does not stand alone. The GIS and open map data guides below take up the questions it leads into:
- QGIS Join Attributes by Field Value: Attach a Table to Your Layer by a Shared Key: A field-value join is a spreadsheet lookup for maps: match rows in a table to features in a layer by a shared key, a code, a name, an ID, and the...
- QGIS Spatial Join: Join Attributes by Location, Explained With Its Predicates: A spatial join answers the question every GIS user meets in the first month: these points know what happened, those polygons know where it counts,...
- How to Add a Shapefile in QGIS: The Whole File Family, the CRS Prompt, and the Zip Trick: Adding a shapefile sounds like opening a file, and it is, once you understand that a shapefile is not a file.
- OpenStreetMap Speed Limits: The maxspeed Tag System and the Defaults Behind It: One key’s complete world: values, implicit codes, sources, and the default tables.
- OpenStreetMap Waterways: The waterway Key Family, from Rivers to Drains: Another key family taught whole: the value ladder, flow regimes, and structures.
Frequently Asked Questions
What is a tag in OpenStreetMap?
A tag is a key=value pair attached to a mapped element, a node, way, or relation. The key names the topic, such as highway or amenity, and the value gives the detail, such as residential or cafe. A feature's full set of tags is its complete description in the database.
Can an element have two values for the same key?
An element carries each key only once. Where several values genuinely apply, the convention is a semicolon-separated list in the single value, such as cuisine=italian;pizza. Separate keys are never duplicated on the same element.
Who decides which keys and values are allowed?
Nobody, centrally. Conventions emerge from community practice, are documented on the OpenStreetMap wiki, sometimes through a formal proposal process, and become real standards when editors use them and software supports them. Taginfo, the usage census, shows which tags the world actually uses.
What does the colon in a key mean?
It separates a key family from a qualifier. Address keys live under addr:, such as addr:street and addr:postcode, and language versions of a name live under name:, such as name:en. The prefix groups related keys so a feature's address or multilingual names read as one dossier.
How many tags does OpenStreetMap have?
By the wiki's figures as of May 2026, the database uses over 100 thousand distinct keys and over 190 million distinct tags. Taginfo tracks the live counts, including how often each key and value is used and where, which is how editors check whether a tag is a living convention.
Photo credits: Kha Ruxury (Pexels); Joaquin Carfagna (Pexels). Photos are served from this site; follow the credit to the library page for the original.