Google September 2026 Spam Update Opens Two-Week Rollout

A spam update, not a core update, sets the tone

The biggest search news on September 25, 2026 came from Google's ranking systems, and it ran opposite to the pattern of the past six months. Google had released its September 2026 spam update, and Search Engine Roundtable stressed that this was a spam update, not a core update, with a rollout of about two weeks instead of the usual two days or so21. The update applies worldwide, covers all languages, and is supposed to affect only sites that break Google's spam policies21. It was Google's fourth spam update of the year23.

The long rollout is what stands out. Google's three earlier spam updates in 2026 each finished in under three days. March took about 19.5 hours, June took about two days, and August took about two days and 16 hours610. The September update started on September 24 and ended on October 8, taking 13 days and roughly 16 hours1014. Google has released one spam update in 2023, three in 2024, one in 2025, and four by September 2026, so this year's pace of spam enforcement is unusual by recent standards9.

Coverage is not fully consistent on the update calendar. Search Engine Journal's algorithm history lists a core update that supposedly ran from September 12 to September 261. Most other trackers built from Google's Search Status Dashboard show only two broad core updates in 2026, in March and May, and nothing in September except the spam update367. Search Engine Roundtable described the September 25 news as "not a core update"21. The more likely reading is that the September core-update entry is an error. Site owners trying to explain September traffic changes should treat the spam update as the confirmed event, alongside unconfirmed volatility around September 23 and 2425.

Why the rollout length matters

A long, staged rollout makes ranking changes harder to diagnose. Search Engine Roundtable reported that the first wave hit hard from Friday, September 25 through Sunday, September 2722. A second phase arrived on September 3024. Found, a UK agency, also pointed to unusually high volatility between September 25 and 272.

No one outside Google knows exactly what the update targeted. Google did not say which spam tactics it was going after2. One analysis noted that Google released no new spam policies alongside the update, which suggests improvements to detection systems such as SpamBrain rather than a change in what counts as spam5. Coalition Technologies argued that heavy abuse of AI-generated content is the most likely focus. It noted that Google has already confirmed that manipulating AI Overviews and AI Mode falls under existing spam policies, but said Google has not confirmed that AI content is a specific target10. Glenn Gabe, posting on the update's first day, said many companies are gaming AI search22. Reporting from earlier in the week supports that view: spammers were using public Gemini Notebook pages to push all kinds of spam into Google results30.

This looks like the central story. The spam war is moving from the ten blue links to the AI answer layer, and Google's quality systems are being retuned for it.

AI Overviews: more links, but more reasons to stay on Google

The same week brought mixed signals on AI Overviews. Data reported on September 23 showed AI Overviews with external links rising from close to 0% to more than 25%26. Search Engine Land reported that AI Overviews now appear on 39% of U.S. desktop searches26. Comscore figures show that share was 25.8% in July 202514.

The extra links do not all point to publishers. Google was testing AI Overview links that send users into AI Mode, its own conversational search product30. Google Discover was testing a "Dive deeper" feature with AI-generated cards that link back to Google, which Search Engine Roundtable said would reduce clicks to publishers3023. Google was also testing a "Loading..." button in place of "Show more" under AI Overviews22. Separately, some AI Overviews now expand automatically into a longer answer that looks much like AI Mode7. Overall, Google is linking out more while also keeping users on its own pages longer.

The publisher traffic numbers keep diverging

The September 25 recap also showed how unsettled the traffic debate is. SEO consultant Barry Adams advised strong skepticism toward Chartbeat data showing publishers' Google traffic down 40% year on year21. Two days earlier, Paul Graham had cited a drop of about 28% in web traffic to news sites over two years26.

The gap between these numbers is real, but every serious estimate points down. A randomized field experiment with 1,065 Chrome users found that showing an AI Overview cut outbound organic clicks by 39.8% and raised zero-click searches by 34.5%13. It also found no improvement in user satisfaction or in the quality of downstream visits13. That result directly challenges Google's claim that clicks from AI Overview pages are "higher quality"12. Ahrefs found a 58% lower click-through rate for top-ranking pages when an AI Overview appears18. Pew found users clicked a traditional result 8% of the time with an AI summary and 15% without one16. A Define Media Group analysis put the drop in organic search clicks at 42% across 64 sites17.

Adams' skepticism of the Chartbeat figure seems reasonable, because aggregate data mixes very different kinds of publishers. Still, both the experimental and the observational evidence point to lower traffic, and the dispute is mostly about how much.

The broader context makes this harder for publishers. On September 30, Judge Amit Mehta dismissed antitrust suits from Penske Media and Chegg. He found the traffic harm real but said it is a matter for Congress, not antitrust law11. Google has confirmed a pilot that pays a small number of publishers, reportedly around 100, for use of their content in AI features11. Since August 31, Google has also offered a worldwide Search Console setting that lets sites opt out of generative AI features. Opting out removes the impressions and traffic those features would send but does not affect regular rankings711.

Answer engine optimization meets a measurement problem

Optimizing for AI answers was a major theme in the day's coverage. Search Engine Land ran pieces on generative engine optimization and on auditing AI visibility without paid tools21. Google's own position is that AI search does not need a separate playbook. A page must be indexed and eligible for a snippet to be cited, there are no extra technical requirements, and Google warns against creating content for every possible query variation4. The September 25 recap also linked to a 70,000-answer, 28-day study on how much AI responses change over time21. That is a reminder that AI visibility is a moving target.

Measurement tools are expanding along with the optimization push. Search Console added a "Multimodal" search-type filter to performance reports that week, splitting web results into text-based and multimodal25. Google also updated its video structured data documentation to support the creator property and to clarify interaction types in interactionStatistic21. The same week, Google's /goto redirect change was described as another obstacle for third-party SEO data21. Google explained the redirects as an anti-abuse measure7.

AI Mode as a category name

Rival products were also in the September 25 news. Microsoft was testing renaming Bing's Copilot search tab to "AI Mode," which is Google's name for its own AI search product21. That small test suggests "AI Mode" may be becoming the generic term for conversational search.

OpenAI had its own quality controversy. Citing 404 Media, Search Engine Roundtable reported that OpenAI uses quality raters, which it calls AI trainers, as Google does, and that some of them cheated by using AI21. The same recap noted ChatGPT Ads expanding to Southeast Asia and Taiwan21, which shows answer engines building ad businesses on top of search-style behavior.

The takeaway

Taken together, the coverage of September 25 shows that the hardest parts of search are moving into the AI layer. That includes spam enforcement, traffic losses, measurement, and branding. The spam update's unusual length and Google's growing push against people gaming AI answers suggest Google sees its generative features as a new target for manipulation. Publishers, meanwhile, are still fighting over how large their losses are while courts decline to step in. Google is adding more links to AI Overviews, but it is also building more ways to keep users in AI Mode. For publishers and SEOs, the practical lessons are to rely on first-party Search Console data, judge traffic changes against the full rollout window from September 24 to October 8, and treat AI visibility as a separate metric that does not guarantee traffic107.