Google finished rolling out its September 2026 spam update on Oct. 8 after roughly 13 days and 16 hours, according to Google's ranking update history and reported this week by Search Engine Journal's SEO Pulse.

How long did the rollout run, and how does it compare?

The rollout started Sept. 24 and ran close to the "up to two weeks" window Google initially cited. All three previous spam updates in 2026 wrapped in under three days, making this one an outlier by a wide margin.

Rank-tracking platform SEOmonitor reported that from Sept. 26, industry-wide rank trackers logged sharp volatility that human searchers did not see. The firm tied that gap to incomplete and inaccurate data Google returned to automated search tools during the rollout.

With the rollout over, SEOs now have a fixed Sept. 24–Oct. 8 window to compare Search Console performance before, during, and after the change. SEOmonitor added that genuine spam updates tend to produce stickier shifts, visible across consecutive days and in Search Console impressions and clicks.

How long does recovery actually take?

Gary Illyes presented timing ranges for crawling, indexing, site moves, and core update recovery on Oct. 2 at Search Central Live Deep Dive Europe in Barcelona. John Campbell of ROAST, one of the event's community speakers, captured the figures in his session recap.

Campbell's recap lists site moves at one to three months in the typical case and six months to more than a year at the slowest. Core update recovery lands at about three to six months typically, and six months to a year in slower scenarios.

The recap does not specify sample size, time period, or what "typical" means. The numbers still give SEOs a yardstick when clients ask whether a site move is dragging past its usual window.

What changed in Search Console?

Google Search Console's Performance report now lets users filter traffic across multiple countries at once. Google Search Central announced the update on LinkedIn on Oct. 6.

The revised help page says users can select multiple countries or devices to see their combined data. The Compare option still shows two devices, countries, Search types, or Search appearances side by side. Users still cannot filter by multiple Search appearances.

For agencies reporting on a portfolio of regions, this cuts a step: one combined Search figure replaces the old two-country compare workflow.

How is Google treating fake bylines?

Google's people-first content guidance now warns against fabricated creator profiles. A new paragraph follows the page's authorship advice and names AI-generated headshots, invented names, and false credentials as forms of deception when they make content look like it came from a human expert. Any form of deception is "a signal of a low-quality page," the guidance reads.

Google's Search Quality Rater Guidelines already classify deliberately misleading creator information as deception. The new language puts that stance in front of site owners reading the helpful-content page directly.

What did Mueller say about AI crawlers?

John Mueller told the Oct. 1 episode of Google's Search Off the Record podcast, alongside Martin Splitt, that AI training crawlers usually offer no way to submit a sitemap.

Mueller suggested a default name such as sitemap.xml or an RSS feed for sites that want those crawlers to find their content. He said he has seen AI training crawlers access his sitemap and RSS files in his own server logs.

For valid sitemaps stuck on "Couldn't fetch," Mueller named two causes: host load and crawl demand, which he tied to a site's perceived quality. Search Console's Sitemaps report help page also lists low crawl demand among the reasons a valid sitemap might fail to fetch.

A sitemap with an unusual name, no robots.txt entry, and only a Google submission is harder for outside systems to locate. AI training crawlers typically lack a console for direct submission. For valid sitemaps showing "Couldn't fetch," both causes sit outside the file itself, by Mueller's account.

What ties the week together?

Each item turns on a label that says less than it appears to. A late-September rank tracker drop may reflect the way Google returned results to automated searches during the rollout, per SEOmonitor. "Typical" in Illyes's timing ranges is the recap's label, with no specific definition attached. A Search Console country figure can now represent a combined total for several markets. Even a byline is a label, and Google now treats fabricated ones used to feign human expertise as deceptive.

For client reporting, the through-line is clear: check what sits behind any number before quoting it.