A torrent file stores metadata used for BitTorrent. It is defined in the BitTorrent specification. Simply, a torrent is data about a target file, though it contains no information about the content of the file. The only data that the torrent holds is information about the location of different pieces of the target file. Torrents work by dividing the target file into small information chunks, found on an unlimited number of different hosts. Through this method, torrents are able to download large files quickly. When a client (the recipient of a target file) has initiated a torrent download, the chunks of target file that are needed can be found easily, based on the data from the torrent itself. Once all the chunks are downloaded the client can assemble them into a usable form. Note: the torrent must be completed before it can be assembled into a usable form. This differs from conventional internet downloads which have only one host and are usable even if the file is incomplete.

A torrent file contains the URLs of many trackers and integrity metadata about all the pieces. It can also contain additional metadata defined in extensions to the BitTorrent specification. These are known as "BitTorrent Enhancement Proposals". Examples of such proposals include metadata for stating who created the torrent, and when.