Two things set it: how much the server can handle without slowing down, and how much the search engine wants to crawl, based on how useful and how fresh the site looks. For a small site it is rarely the limit. For a catalogue that can generate thousands of URLs, it starts to matter.
The budget is spent on every URL the crawler finds, useful or not. Thin pages, duplicate addresses, redirect chains and dead links in the sitemap all take fetches that could have gone to the pages you care about. So the work is mostly subtraction: fewer URLs that do not deserve a visit, and clearer paths to the ones that do.
In practice that means a sitemap with only indexable URLs, internal links from the body of related pages to the pages that matter, a clear status code for removed pages, and no parameter variants that multiply one page into many. Search Console does not report a single budget number. The coverage report and the crawl stats are where its effects show.
An example from our own work
We have not measured Videojam’s crawl budget as a number, so we describe its shape instead. Videojam, our own video platform, is a catalogue of videos, artists, genres and events that can generate thousands of mostly thin URLs. Its indexation gate keeps thin pages out of the sitemap, and the home and DJ-sets pages carry server-rendered links to the current chart and playlists. The coverage report dated 23 October 2026 reads as a funnel: 2,445 pages known to Google, 2,068 crawled, 970 indexed. 377 were discovered but not yet indexed. On audacian.eu, removed pages answer 410 Gone rather than 404, so crawlers can drop them faster.