What Google's August 2026 Spam Update Actually Hit

Google ran three spam updates in 2026: March, June, and August. The August update, which finished rolling out on August 21, knocked more top 10 URLs out of the top 100 than either of the other two, going by SE Ranking's data. The case studies published since show scaled, AI-assisted content took the worst of it.

If your traffic fell off a cliff sometime between August 18 and 21 this year, you're not imagining it. Search Engine Land pulled data from SE Ranking for the five days on either side of that window and found something that doesn't usually happen with a routine spam update: 16.71% of URLs holding a top 10 spot fell clean out of Google's top 100 for the same keyword, against 9.2% during a comparable stretch in July with no confirmed update running. Put another way, a page sitting in the top 10 that week was roughly 1.8 times more likely to disappear from the top 100 than it would have been in an ordinary one.

A few days after the rollout finished, Glenn Gabe at GSQi set out to find who actually got hit and published four case studies with the query counts attached. They're the closest thing to ground truth anyone outside Google has for this update, and they all lean the same way: sites publishing at scale, with little in each page beyond what a template or a model could produce, took the worst of it.

Even if you don't run anything close to a content farm, one change from earlier in the year matters more than the update itself. Back in May, after the March update but weeks before the June one, Google quietly extended the same spam rules to cover what its own AI Overviews and AI Mode say about you. Since then there hasn't been a separate rulebook for ranking spam and AI answer spam, it's one set of policies covering both, and that's the part I'd pay attention to even if the August numbers never touched you.

So what follows is the SE Ranking data first, then Gabe's four cases, then the three checks I'd run on your own site before deciding you're clear.

Three spam updates in 2026, and the slow one did the damage

Line the three up side by side and the first thing you notice is that rollout speed told you nothing about how hard an update would land. The one that took longest to finish is the one that moved the most rankings.

None of the three came with a new named policy or a blog post explaining what changed. Google's spam updates documentation hadn't been touched since December going into the August rollout, and Barry Schwartz at Search Engine Roundtable confirmed directly with Google that this update left link spam and site reputation abuse alone. So whatever changed, changed inside SpamBrain, Google's existing AI-based spam detection system, and it was enforcing policies already on the books rather than anything new.

Why the August update moved rankings harder than usual

SE Ranking ran its comparison across the same 100,000 US keywords in 20 industries, checking results from August 17 against August 22, then set that against a baseline from July 26 to July 31, a week with no confirmed update running.

Worth being clear that this is a different measure from the 16.71% figure above. That one counted pages falling out of the top 100 altogether; this one is top 10 volatility, how much the top 10 itself reshuffled, and every one of the 20 industries saw more of it than the baseline week did. Real estate saw the least, at 74.64%; fashion and beauty saw the most, at 85.55%.

The categories Google tends to move most cautiously on, real estate and healthcare among them, were the steadiest of the 20. That tracks with how conservatively Google usually treats YMYL content, the pages where being wrong costs a reader money or their health, though that's my read of the pattern rather than anything Google has said.

What that data can't tell you is which pages lost ground or what replaced them. SE Ranking's analysis sits at the keyword level, not the URL or domain level, so it measures how much the results reshuffled without saying why. For that you need someone who opened up individual sites one at a time, which is what Gabe did next.

The rule that now reaches AI Overviews and AI Mode too

After the March update shipped, and six weeks before the June one, Google made a change to its Search Central documentation that matters more for anyone doing AI visibility work than anything in the August numbers. On May 15, 2026, Google added a line to its spam policy naming attempts to manipulate generative AI responses in Google Search as spam, sitting right next to the older line about manipulating rankings. Google was explicit that this wasn't a new policy. The existing ones (cloaking, scaled content abuse, link spam, site reputation abuse, doorway pages, scraping, thin affiliation) now reach AI Overviews and AI Mode as well as the ranked list, and nothing about how they're enforced changed along the way.

Gabe made the same point from the other direction in his case studies. When a spam update takes a site down, the loss isn't confined to the 10 blue links. It can show up in AI Overviews, in AI Mode, and even inside ChatGPT's answers on the occasions ChatGPT is grounding its response through Google. There's no such thing as being partially penalised anymore; one finding follows the site across every surface Google's index feeds, including the ones you might not have realised you were competing on.

What Glenn Gabe actually found when he went looking

Gabe doesn't name the sites in his case studies, which is standard practice for him and fair enough, but he does publish the query counts, and they're specific enough to be useful.

Link spam wasn't what the August update was after, Google confirmed that much directly, and every one of Gabe's four cases bears it out. They're all the same kind of content-side problem: a lot of pages that exist because a template or a product feed could produce them, with nothing on them anyone actually asked for.

The three policies almost every hit maps back to

All four of Gabe's cases, and most of what turns up in spam update case studies generally, trace back to one of three policies Google named in March 2024 and has been enforcing ever since.

The detail that trips people up is that scaled content abuse doesn't care how the pages got made. Google says as much directly: a large site can publish thousands of pages and stay completely fine if each one adds something real, while a small site with a few dozen thin, templated pages can still be in breach. The page count on its own was never the violation. What Google is trying to judge is whether a page exists for a reader or for a ranking, and volume is just one of the signals it uses to guess at that.

If you're not running a content farm, does any of this apply to you

Almost nobody reading this runs 250,000 programmatic URLs in a risky niche, so the honest answer is probably not, at least not directly. I'd still run these three checks before assuming you're clear.

None of this is really about AI content specifically, whatever the headlines around it suggest. Google has said the same thing since 2023: it judges quality and originality, not who or what typed the words. What changed this year is that the judgement now follows a brand into AI Overviews, AI Mode, and whatever ChatGPT borrows from Google along the way.

It's the same thing I keep running into when I look at why good SEO on its own doesn't guarantee AI citations, and why AI models skip brands they can't clearly place in the first place. It's no longer one core strategy that works for everything, it's a lot of small ones that together add up to visibility, and the ones that have worked for us were built one page at a time rather than run off a topic list. That's why our approach to generative engine optimisation at BrandViz.AI stays surgical, not mass, even when a client wants 30 pages this month instead of 3 good ones.

Common questions about the August 2026 spam update

Was the August 2026 update only about AI-generated content?

No. Google didn't name a target for this update at all, and confirmed separately that it left link spam and site reputation abuse untouched this time. The published case studies show a concentration of scaled content abuse and thin affiliate setups, some AI-written, some not. Content made mainly to rank, whatever produced it, is what keeps showing up.

How do I know if a drop in my traffic was actually caused by this update?

Check Google Search Console's Manual Actions report first, since a spam update almost never triggers one. Then compare when your decline started against the confirmed rollout window of August 18 to 21, 2026. If the timing lines up and the affected pages share a thin, templated, or heavily automated shape, that's the pattern spam updates target. If your decline predates the window, look elsewhere first.

Can a site recover after being hit by a spam update?

Yes, according to both Google and Glenn Gabe's own tracking of past recoveries, but only after the offending pages are removed or substantially rewritten rather than lightly edited, and only after Google observes a clean site over an extended period, usually months. Sites built entirely on thin, programmatic content with no first-party value underneath face a harder path, since there's nothing left to improve toward.

Do Google's spam rules for AI Overviews and AI Mode apply to my brand even if I don't target them directly?

Yes. Since May 15, 2026, Google's existing spam policies formally cover generative AI responses in Search, including AI Overviews and AI Mode, using the same enforcement as ranking spam. A finding against a site under any existing policy, scaled content abuse or cloaking among them, can now cost visibility on those AI surfaces as well as the ranked list, whether or not the site ever tried to target them specifically.