How crawl budget works
Two separate limits combine into one. The first is what your server can comfortably take: Google watches response times and error rates while it works, and eases off when a site starts to struggle. The second is how much crawling your content is felt to deserve. Pages that change often, attract links and get visited are revisited readily; pages that never change and nobody points to are fetched rarely, and eventually barely at all.
Together they set a rough ceiling on how many of your addresses get fetched in a given period. Google’s own guidance is blunt about who should care: most sites should not. It becomes a genuine constraint on sites holding an enormous number of URLs, or on sites that manufacture addresses faster than anyone can count them — filter and sort combinations, internal search results, calendars that run forward without end.
Why crawl budget matters
Everything else waits on crawling. If most of the fetching is spent on filtered listing pages nobody should ever see, newly published products and freshly updated pages sit in the queue behind them. On a large retail site that appears as a lag between publishing something and it showing up in search at all.
It is also worth reading as a symptom. A crawl rate that falls while the site keeps growing usually points at server speed rather than at lost interest. A crawl rate spread thinly across thousands of near-identical addresses points at an architecture problem that is costing you in other ways as well.
Common mistakes with crawl budget
Worrying about it on a small site is the most common. A brochure site or a local service business with a modest page count will never approach a ceiling, and the hours spent here are hours not spent on content or conversion. Look at the crawl stats report before deciding you have a problem.
Confusing crawling with indexing is the next. A page can be crawled and still left out of the index, and a page blocked from crawling can still surface in results when enough links point at it — search engines simply have no content to show alongside it. Blocking an address does not remove it, and it also prevents anyone reading the noindex instruction sitting on the page.
The related error is treating noindex as a way to save crawling. The crawler has to fetch the page to read the instruction, so it saves room in the index, not requests.
How to act on it
Start with the crawl stats report in Search Console, which shows how much is being fetched, how quickly the server responds and what those requests are being spent on. Slow responses are the first thing to fix, because they lower the ceiling directly and hurt visitors at the same time.
Then cut the number of addresses worth fetching. Close off the endless spaces created by faceted navigation, internal search and open-ended calendars. Consolidate duplicates onto one chosen address. Keep the sitemap honest, listing only indexable URLs with truthful modification dates. Make sure the pages that actually earn money sit within a few clicks of the home page, because crawlers follow links much as people do. That structural work belongs to technical SEO, and it pays off whether or not a budget ceiling was ever your problem.