Google’s explanation for this year’s run of spam updates is a supply problem: the web is producing far more new pages than before, and a large share of them is spam. That account comes from We Are ROAST’s recap of the final day of Google Search Central Live Deep Dive in Barcelona, whose final day was October 2. The recap is by John Campbell. Search Engine Roundtable’s Barry Schwartz surfaced it on October 5.
Google has not published the slides. Everything below is a secondhand recap of live talks, so treat the wording as ROAST’s rendering, not Google documentation.
According to ROAST, Duy Nguyen, a Search Quality Analyst at Google, covered spam in his quality session. The recap says a high share of web content is spam, that this is what spam updates target, and that the abundance of new content explains the frequency of recent updates. It also says Google’s team now has started applying AI to assist in detecting more of that spam. ROAST does not say how the AI is applied or how much it changes detection rates.
Two other points came from Eric Murillo, a Trust and Safety Analyst for Discover at Google. Per the recap, he said AI slop is growing, which puts more weak pages into the Discover feed. He also said user feedback matters most here, because it is the raw material for Google’s newly built spam filters.
Gary Illyes ended the morning session on a single figure: Google removes roughly 40 billion spam pages every day. Schwartz notes the figure is not new, since Google shared it well before this event. Treat it as context for scale, not as fresh disclosure.
The more useful material is Illyes’ three reasons Google keeps revising how it judges quality, as ROAST recorded them:
- Formats: Search now spans far more media than ten blue links, and each format needs its own quality judgment.
- Breadth: The index holds many more pages and sites. Core updates evaluate individual pages, not whole domains, which is why some URLs on one site rise while others fall.
- Spam: Updates here sharpen detection of cloaking, doorway pages, scraped content, hacked content, and link spam.
Illyes also warned that mass-produced content is now a larger problem than link spam. For years, spam work in many agencies centered on link profiles. A team publishing hundreds or thousands of templated pages should now treat page-level usefulness as the exposure to audit, because a template that serves one hundred pages well may offer little at ten thousand.
Timing matters for anyone reading current volatility. Google’s September 2026 spam update began September 24, and as of October 6 the status dashboard shows no completion. Our earlier report on the second wave on September 30 covers that rollout. Nothing in the recap links the AI-assisted detection to any specific site’s movement, so a ranking drop during an unfinished update is not evidence that the detection system flagged you.
Some of the same event’s material is covered separately. Illyes’ crawling, indexing, and recovery timings come from this same event, and Google’s people-first content page recently changed on main content quality and fabricated authors.
The recap offers no independent measurement of how many pages the AI step catches, and Google has not said whether new thresholds are coming. ROAST’s own takeaway is to expect more updates and tighter thresholds, which is Campbell’s reading and not a Google statement.
Before the September update closes, teams publishing at scale should sample their templated sections for pages that add nothing beyond the template, and hold major content changes until the dashboard shows completion so movement can be attributed cleanly.
Reported by Search Engine Roundtable (Barry Schwartz, October 5, 2026), based on We Are ROAST’s recap by John Campbell of Google Search Central Live Deep Dive Barcelona, Day 3 (October 2, 2026).
