Google September 2026 Spam Update Complete: 13 Days, 16 Hours

On 24.09.2026, shortly after the announcement, I opened Google’s ranking release history. At the top was a new row with the status “Information” and no duration. Same as in August. What was different was the sentence in the entry: “up to two weeks”. In March, June and August the same spot said “a few days”. Since 08.10.2026 it shows a duration: 13 days, 16 hours.

I had already covered three spam updates here this year. None of them ran longer than three days. Compared with the three announcements before it, only half a sentence has changed. That half sentence is the news. Whether it means anything can be checked against Google’s own list. That is what I did. I opened all twelve spam update entries in the release history and put each announcement next to the measured runtime.

As with the August 2026 spam update, this article separates what is confirmed, what remains open and the question of how you tell an update effect from a measurement effect. For the last part there is a reason this time that showed up in Search Console on the same day.

The entry in the Google Search Status Dashboard reads verbatim: “Released the September 2026 spam update, which applies globally and to all languages. The rollout may take up to two weeks to complete.” The header of the same page now gives start and end: “Incident began at 2026-09-24 09:15 and ended at 2026-10-08 01:00 (all times are US/Pacific)”. That is 18:15 CEST on 24.09.2026 and 10:00 CEST on 08.10.2026. Below it, since 08.10.2026 at 01:37 Pacific time, there is a second line: “The rollout was complete as of October 8, 2026.”

On LinkedIn, Google Search Central worded the announcement in a little more detail on the same day: “This is a normal spam update, and it will roll out for all languages and locations. The rollout may take up to two weeks to complete.” The same phrase “a normal spam update” was also in the LinkedIn announcement of the August update. There Google also promised a follow-up: “We’ll update our ranking release history page when the rollout is complete.” Google delivered it on 08.10.2026 and on the same morning also reported the completion in a comment under the post and on X.

It is the fourth spam update Google lists in its release history this year, after March, June and August. The gaps from start to start have become shorter: 92 days from March to June, 55 days from June to August and now 37 days. In the release history, which goes back to November 2021, no other calendar year has more than three entries that Google labels as a spam update.

The list only shows what Google has entered there. Why the gaps are getting shorter is not stated anywhere. I am not going to guess.

The announcement names no target, and neither does the completion entry. Neither does the documentation page on spam updates, which on 08.10.2026 still carries the date of 10.12.2025. What it says is general. For affected sites: “Sites that see a change after a spam update should review our spam policies to ensure they are complying with those.”

Then there is the timeline for a recovery: “Making changes may help a site improve if our automated systems learn over a period of months that the site complies with our spam policies.” Months, not days. And for the special case of link spam, the page states explicitly that a fix may bring no improvement, because an earlier benefit from devalued links is lost.

The second gap is new. Google does not say why this update was announced with two weeks while the three before it were announced with a few days. Asked by Barry Schwartz whether it really should say two weeks, John Mueller replied on the start day on Bluesky: “Yes, it’s likely to take longer than some of the previous ones.” He gave no reason. A longer duration can mean many things. None of them is confirmed.

In the Status Dashboard each of the twelve spam updates so far has its own entry with the wording of the announcement. After completion, the release history shows the total runtime. I read both on 24.09.2026 and added the September row on 08.10.2026:

Source: Google Search Status Dashboard, ranking release history and the individual incident entries, read on 24.09.2026, the September row on 08.10.2026. The announcements are quoted in Google’s wording, for March 2024 including the missing “to”. Google only added the three entries from 2021 and 2022 in March 2023, without an expected duration. Linked are the updates that have their own article here.

Two things stand out. First, the announcement roughly got the order of magnitude right every time. A few days turned into hours or two to three days. A few weeks turned into almost four. Second, in these three cases “up to” was not a ceiling accurate to the hour. In all three cases the duration Google reported was slightly longer, by one hour in June 2024, by two in December 2024 and by 21 hours in March 2024.

The release history has exactly one earlier spam update announcement with two weeks, the one from March 2024. It is not directly comparable, because the March 2024 core update started on the same day. Movement from those weeks therefore cannot be cleanly assigned to either update.

The September update turned out differently. Two weeks from the start would have ended on 08.10.2026 at 09:15 Pacific time, which is 18:15 CEST. The rollout was already done at 01:00 Pacific time, a good eight hours earlier. Of the four “up to” cases in the release history, it is the first that came in under the stated duration. For planning, in my view, this means: the order of magnitude has been reliable so far, the hour has not. That is an observation across eight cases and not a rule.

The last spam update that ran longer than this one was the August 2025 update, at 26 days and 15 hours. The August 2026 update took only about a fifth as long. Of all twelve spam updates in the release history, four ran longer than the September one. The median of all twelve is now 7 days and 13.5 hours.

On 08.10.2026 I pulled the daily values of two trackers for Germany: the Semrush Sensor across all categories and the Sistrix Google Update Radar, which only measures mobile. Each evaluates its own keyword set on its own scale. Semrush classifies values from 5 to 8 as high and from 8 to 10 as very high. Sistrix reads values above 6 as a relevant update.

Source: Semrush Sensor (Germany, all categories, scale 0 to 10) and Sistrix Google Update Radar (Germany, mobile, index value with an average of 1 over 90 days), retrieved on 08.10.2026. The values for 08.10.2026 are daily readings at the time of retrieval.

I read three things from this. First, Germany came later than the US. Semrush’s desktop value for Germany stayed at 0.6 until 25.09.2026, while for the US it already stood at 9.4 that day. Second, desktop moved more than mobile. On desktop, Semrush showed values above 8 on four days; on mobile the sensor stayed below 6 throughout the rollout. Third, the three phases that Search Engine Roundtable described for the US trackers can be recognised in the German Semrush values. Sistrix does not show the middle one around 01.10.2026.

By far the highest value is Sistrix’s 136.34 on 27.09.2026. For exactly these days, several providers report that querying the search results was disrupted. On 26.09.2026 Sistrix wrote on its status page that Google was in some cases returning incomplete or distorted search results. SEOmonitor describes that from 26.09.2026 Google sent incomplete and incorrect results back to automated searches. In rank trackers this looked like sudden drops or spikes; people searching themselves did not see them. SEOmonitor therefore carries over the values of 25.09.2026 for 26.09.2026 and 27.09.2026. So part of the spikes in the table may come from the measurement and not from the update. Which part cannot be separated from the outside.

Diib published a first analysis with visit data on 29.09.2026. It compared the days from 24.09.2026 to 26.09.2026 with the same days one week earlier, across 32,569 websites with connected Google Analytics. The median share of visits from unpaid Google search fell from 34 to 31 percent. For 10 percent of the websites it halved or fell further, compared with 5 percent in the comparison week. Diib itself calls this an early read, not a verdict. The data shows neither which websites Google classified as spam nor how things developed after the first three days.

For seo-kreativ.de, Search Console shows no sign of a hit up to 06.10.2026. The week from 30.09.2026 to 06.10.2026 brought 259 clicks, the week before the start 257. Of 341 queries with at least 10 impressions in both weeks, none dropped from page 1 to a position beyond 20. The manual actions report shows no issues. This is an interim reading with the Web (text) search type. I will add the measurement using the method from my guide as soon as the week from 16.10.2026 to 22.10.2026 is fully available in Search Console.

As in August, two things were lumped together with the update. Each of them fits either not in time or not in substance.

First: the spikes from 22 or 23.09.2026

On 24.09.2026, Search Engine Roundtable reported significant spikes across several trackers, starting around 23.09.2026, maybe as early as 22.09.2026. The spam update announcement only came after that. According to Google, the update started on 24.09.2026 at 09:15 Pacific time.

On the question of whether Google rolls out earlier than announced, Search Engine Roundtable reports a reply by John Mueller under the August announcement: “We don’t roll them out beforehand. We try to get the announcement as close as possible to the actual “button-pushing” (which is sometimes challenging due to time-zones).”

I am not saying what caused the spikes. I am only saying: going by the start time in the Status Dashboard, this rollout had not begun at that point.

For Germany I pulled up the Semrush Sensor for mobile on 24.09.2026. It looks different there. Across all categories the value stood at 1.6 on 22.09.2026 and on 23.09.2026 and at 1.7 out of 10 on 24.09.2026. Semrush classifies that as “Low”. The highest value before the start is 4.9 on 19.09.2026. Semrush only classifies that one as “Normal” as well. The screenshot below was taken on 08.10.2026 and also shows the course afterwards.

Two caveats belong with this. Semrush itself shows a notice that the formula for baseline volatility was recalculated on 18.09.2026 and that this may have changed the scores. I therefore do not read a comparison across that day as a finding. And according to its own help page, Semrush checks a fixed set of keywords daily and compares with the previous day at the end of the day. The time and time zone are not stated there. Whether the value for 24.09.2026 already includes the rollout start at 18:15 CEST therefore cannot be read off.

Second: Web in Search Console now means Web (text)

On 24.09.2026, Google announced a new search type on the Search Central Blog. In the performance report, Web now has two sub-options, text and multimodal. Multimodal covers searches where an image was part of the query, for example via Google Lens or Circle to Search.

At midday on 24.09.2026 I measured the new search type in the seo-kreativ.de property. Two findings from that matter for any before-and-after comparison. The report opened for me with “Web (text)”, which is only part of Web. And in my request on 24.09.2026, the Search Console API with type=web returned exactly the values of “Web (text)” for this property, without multimodal data. I have not checked other properties.

For a spam update that is unfortunate, because both happened on the same day. Whether exports you pulled under “Web” before 24.09.2026 included multimodal searches, I cannot say. In my property the multimodal series starts on 10.09.2026. Comparing an old export with a new one may therefore also measure the switch.

I am framing this as conditions, not as a recipe. What follows is my assessment and not a prediction about your rankings.

1. Save the before window. My rule of thumb: the four weeks up to and including 21.09.2026. That keeps the spikes from 22.09.2026 and the start day outside.
2. Settle on the search type. Choose “Web (text)” or text plus multimodal and stick with it for both windows. In my request the API returned only text for seo-kreativ.de. Check this for your own property before you compare API and interface values.
3. Start the after window on 16.10.2026. The release history has shown the runtime since 08.10.2026. For core updates, Google recommends waiting at least one more full week after that. That is written for core updates, not for spam updates. I still think it is a sensible minimum. A one-week after window therefore runs from 16.10.2026 to 22.10.2026 at the earliest. My guide to Google core updates shows what that looks like with real data.
4. Read volatility in your own context. Trackers such as the Semrush Sensor show unrest across the SERPs, not in your own data. A spike only stands out once it leaves your usual range of fluctuation.
5. Go through the spam policies if you are affected. That is Google’s own recommendation. It comes with the time frame “over a period of months” and not with a deadline in days.

What I do not recommend: large-scale changes to content or links while the target of the update is not known. A rebuild based on a suspicion is at best ineffective.

The September 2026 spam update ran from 24.09.2026 to 08.10.2026, 13 days and 16 hours. It was announced with “up to two weeks” instead of “a few days” as before. Google got the order of magnitude right again. This time the rollout was even done a little earlier than stated.

More important than the update itself is the question of whether what you see in your data comes from it at all. Two candidates speak against it: the spikes from 22 or 23.09.2026 came before the start. And on the same day, what “Web” means in Search Console changed.

I will add the full measurement for seo-kreativ.de as soon as the after window is available in Search Console. If Google says which policies the update addressed, that will be added here as well.