On 24 September 2026, Google launched its fourth spam update of the year, applying to every country and every language. Its status dashboard announces a rollout of up to two weeks, so a possible end around 8 October, and specifies neither the target nor the scale of the update. As of 1 October, the incident page shows only the launch entry. For a marketing team, the question of the week is how to read a traffic drop during this period without jumping to conclusions and without touching pages that were working. The observations published over the past week give leads, which no data from Google confirms at this stage.
The facts
- Google announced the update on 24 September at 9:15 am Pacific time on its Search Status Dashboard: it applies globally and to all languages, and the rollout may take up to two weeks to complete. According to Google's documentation, a spam update improves its automated spam detection systems, including SpamBrain, and an affected site should review the spam policies; an improvement, if it comes, takes several months.
- The three previous updates of 2026 were much shorter according to the dashboard: about 19.5 hours in March, a little over two days in June, about two and a half days in August. PPC Land notes that Google Search Central called the September one a normal spam update on LinkedIn, with no blog post, no new policy and no list of targeted practices.
- On launch day, Google's John Mueller replied on Bluesky that it would probably take longer than the previous ones, Semrush reported on 29 September. Search Engine Roundtable recorded strong volatility from 25 to 27 September, peaking on Saturday 26, on the fifth day of a rollout announced for fourteen.
Google confirmed the launch and the duration, and left the rest open
Google's communication fits in two sentences. It confirms the launch, the global scope and a rollout of up to two weeks. It points, as with every spam update, to the spam policies that affected sites should review. Nothing is said about the policy concerned, the type of site targeted, the share of sites affected or the end date.
Two hypotheses are circulating without confirmation. Search Engine Journal described on 25 September a large-scale abuse detection system called SAFE, presented by Google in a research paper, and suggests it could be part of the September update; Google has not said so. Semrush notes that this spam update is the first confirmed since the revision of the site reputation policy on 28 August, without Google having linked the two. PPC Land lists the existing policies the update would enforce, from scaled content to fake reviews: that is the list of rules in force, which Google did not present as September's target.
Three things blur the reading of a drop during the rollout
The duration first. A drop seen on the fifth day of a two-week rollout says little about the final state: positions can still move in both directions until the close, and Search Engine Roundtable and Semrush both point out that this calendar makes it harder to attribute a drop to the update.
The volatility next. Several tracking tools recorded movement from 22 September, before the announcement, according to PPC Land. A comparison that starts on 24 September therefore mixes two periods; Semrush recommends taking positions prior to 23 September as the baseline.
Search Console last. On the same 24 September, Google split the Web search type in two in the performance report: text-based search and multimodal search, started from an image with Google Lens or Circle to Search. Google presents this multimodal data as new, and no query is attached to it since the person searched with an image. A rise in impressions seen since that date may therefore come from the report itself; clicks and conversions remain the measure least exposed to this effect.
Pages produced in bulk come up in every observation, with no overall measurement
The available observations come from consultants and cover limited samples. In Search Engine Roundtable's article of 28 September, Glenn Gabe reports big drops across a number of sites, in several verticals and countries, with cases of AI-generated content and programmatic pages at scale; Lily Ray describes highly templated pages, varied by phone number or area code. These are isolated cases, with no overall measurement to date.
The August update gives an order of magnitude. According to SE Ranking, whose analysis is relayed by seo-kreativ in an addendum of 31 August, 16.71% of URLs ranked in the top 10 fell beyond position 100 during the update, against 9.2% over a July window with no update, or 82% more movement. The study covers 100,000 keywords and 20 industries on US results, and does not carry over to September. In his August case studies, Glenn Gabe describes sites that combined programmatic pages, AI-generated text and thin affiliation; DMNews recalled on 3 September that these observations rested on a small sample and that Google had not confirmed a target.
One point converges: sites that produce pages in bulk would be among the hardest hit. Google's official definition of scaled content abuse sheds light on this. It targets pages generated in large numbers to manipulate rankings without helping users, no matter how they are created, and cites among its examples the use of generative AI tools to generate many pages without adding value. The criterion is the value delivered, and the production method does not count. AI-assisted content that is proofread, enriched with first-party data and signed by an identified author follows a different logic, because a human keeps editorial control.
In the European Economic Area, site reputation is now handled separately
On 28 August, Google published an update to its site reputation policy, after exchanges with the European Commission. This policy targets third-party content published on a strong domain to benefit from its ranking signals; wire stories, syndication and properly disclosed affiliation remain allowed. Since 30 August, a manual action for site reputation abuse no longer has any effect for users located in the European Economic Area, which brings together the 27 EU countries, Iceland, Norway and Liechtenstein. In that area, the pages concerned can be ranked separately from the domain hosting them, on their own merits, and Google is lifting past manual actions. Elsewhere, the manual action continues to reduce the visibility of the section concerned.
Google's documentation does not explain how this treatment combines with the September spam update. A site hosting third-party content can therefore get different results depending on where users are located. On 29 September, Nick Fox, Google's Senior Vice President for Knowledge & Information, the division that includes Search, said in an interview relayed by Search Engine Roundtable that Google had been forced to pull back on spam protections and was showing lower-quality results in Europe as a result. We described this framework in our article on the suspension of parasite SEO penalties in Europe; the September update is the first to apply within it.
What to do this week, and what is better avoided
A drop observed during the rollout can have several causes, hence the value of diagnosing before acting. These checks follow Google's documentation and Semrush's note of 29 September, with no guarantee of results.
- Wait for the close before concluding. Google's dashboard will show the end of the rollout, possibly around 8 October.
- Set a baseline. Compare positions and traffic before 22 September and after 24 September, since volatility began before the announcement.
- Look at competitors. If the whole sector is moving, the cause goes beyond your site. If yours is the only one moving, examine its pages.
- Check outside Google. Semrush suggests comparing with your citations in ChatGPT, Gemini or Perplexity. A drop limited to Google, while those citations hold, points to a Google-specific cause. It is one of the indicators our GEO service tracks.
- Isolate the Search Console effect. Since 24 September, Web search is split there into text and multimodal. Track clicks and conversions, and read impressions with that change in mind.
- Check manual actions and messages in Search Console, remembering that a spam update changes automated systems and triggers no individual notification.
- Take stock of pages at risk. City or product pages replicated at scale, automated texts never proofread, affiliate listings copied from the merchant, third-party sections, purchased domains. Once the rollout is over, decide batch by batch: enrich, merge, remove from the index or delete.
| What you see | Cautious reading | Action |
|---|---|---|
| General drop across the sector | Broad volatility | Wait for the end of the rollout |
| Drop on a batch of pages generated in series | Consistent with published observations, without proof | Audit the added value of the batch |
| Drop on a section of third-party content | Possible link with site reputation | Check manual actions, compare inside and outside the European Economic Area |
| Rise in impressions with no change in clicks | Possible effect of the new Search Console report | Track clicks and conversions |
| No drop | Site spared at this stage | Monitor until the announced close |
Three moves are better avoided during the rollout: rewriting or deleting pages in bulk; relaunching mass page production with AI tools to make up for the drop, which matches the official definition of scaled content abuse; expecting a quick recovery, since Google says an improvement, when it comes, shows over several months.
AIxH's view
We produce our content with AI, and we publish two to three articles a week on this site: we are therefore concerned by the definition of scaled content abuse, and we have read it closely. What sets us apart from it comes down to verifiable choices on every page: open and linked sources, figures tied to their scope, an identified author who proofreads and signs, and an answer to a question our clients ask. For the sites we follow, we set the baseline at 22 September and we are waiting for the close before any decision on a page. If one of your batches of pages has lost visibility since 24 September, our organic search team can examine its added value page by page, and compare what you see in Google with your citations in AI answer engines, to tell an effect of the update from a content problem.
