tarpit action (alias: poison) never tells a bot it is blocked. Matching requests get a normal 200 OK page of deterministic generated garbage, so the same URL always shows the same fake article, but every URL is different. Each page links to further garbage pages, so scrapers that follow links wander an infinite maze instead of hammering your upstream or learning they have been caught.
The tarpit feature is for times when you’d rather not tip off that they are blocked and instead results that get potential scarped are poisonous. Humans should never match a tarpit rule because unlike deny there is no obvious signal that anything is wrong, so be cautious.
Two optional settings:
tarpit.seed— extra entropy mixed into the generation seed. Defaults to yourcookie_secret.tarpit.corpus— path to a plain-text file used as the source material (any book-length text works). Omit it to use the built-in corpus.