A documentation polish is not a crawl-algorithm launch

On 22 July 2026 Google recorded a change to the public guide Optimize your crawl budget. The change is real. It is also easy to misread. The publisher described a documentation polish. The public HTML did change. Neither fact is a stated change to crawl scheduling or ranking.

Fact. The crawling-infrastructure changelog for 22 July 2026 says, in its What field, that Google polished and clarified that guide to improve clarity, terminology consistency, and flow. The Why field is narrower still: make the documentation easier to understand, and use technical terms such as crawl capacity limit consistently. The matching RSS item has guid 2026-07-22-Polished-and-clarified-the-crawl-budget-guide and pubDate Wed, 22 Jul 2026 00:00:00 +0000. The live guide footer reads Last updated 2026-07-22 UTC. The HTTP Last-Modified header on the live URL, retrieved 24 August 2026, is Wed, 22 Jul 2026 18:28:48 GMT. None of those fields name a ranking system, a crawl-rate formula, or a launch. Changelog: https://developers.google.com/crawling/docs/changelog RSS: https://developers.google.com/crawling/docs/changelog/crawling_docs_updates.rss Live guide: https://developers.google.com/crawling/docs/crawl-budget

Fact. A Wayback Machine capture of the same path from 15 July 2026 06:16:11 UTC supplies a pre-edit text. A row-level comparison against the live page yields thirteen coded units: three new sentences, three wording edits, two terminology refreshes (Index Coverage → Page Indexing), and one each of heading rename, definition rewrite, operational detail, scope-and-wording, and new sub-bullets. The section “How do I get more crawl budget?” is unchanged in substance. Wayback: https://web.archive.org/web/20260715061611/https://developers.google.com/crawling/docs/crawl-budget

Fact. Three sentences in the 24 August 2026 live text have no counterpart in that 15 July capture. One states that for Google Search, not every crawled page is necessarily indexed; after crawling, a page must be evaluated, consolidated, and assessed. One states that every site starts with the same default, conservative crawl capacity limit, which systems may raise if demand exists and the site stays healthy. One states that crawl demand differs by crawler, but the crawl capacity limit is shared, so high demand from one crawler can reduce capacity for others.

Fact. The Search Central “Latest documentation updates” list for July 2026 includes other items (for example 29 July 2026, analyzing social and video platform content in Search Console). It does not list this crawl-budget polish. That is consistent with the late-2025 split of crawling-infrastructure docs onto their own changelog. On 18 December 2025 that same crawling changelog recorded the migration of this guide and wrote that functionality had not changed, only location and some wording. That sentence is about 2025, not about 22 July 2026. It is still Google’s own language that a docs move need not be a crawler-behavior change. Search Central updates: https://developers.google.com/search/updates

Interpretation. The useful object here is a category boundary in indexing-first retrieval. An index is built from what crawlers are allowed to fetch under a capacity policy. Public documentation of that policy can be rewritten without any claim that the policy engine changed. Treating 22 July 2026 as an algorithm date collapses publisher meta-claim (clarity, term consistency) into systems inference. The new public sentences are interesting as eligibility language: crawl does not entail index; capacity is a shared host-level pool; the default limit is conservative. They still only prove what Google published.

Limit. One Wayback snapshot is not a version history. Intermediate drafts are invisible. “Last updated” is a page-level stamp, not a per-sentence authorship date. HTTP Last-Modified is document metadata. No crawl logs, CrUX, or Search Console time series. Do not attribute site-level crawl or ranking movement to this date. Newly published sentences are compatible with documenting long-standing behavior, documenting a prior unpublished change, or documenting a same-day systems change. The retrieved record cannot distinguish those three.

Takeaway. If you work on crawl–index–serve, date the documentation event and keep it separate from any systems event you have not been shown. Shared capacity and crawl-versus-index now sit in the public guide. That is a documentation fact. It is not, on these sources, a crawl-algorithm launch.

Mikhail Drozdov is an AI Search & Indexing Systems Researcher. Casinokrisa is the research platform. Person URL: https://casinokrisa.com/person/mikhail-drozdov