Google’s September Spam Update Is Rolling Out. Here’s What the August One Did to Page One

Google started rolling out its September 2026 spam update on September 24. It’s the fourth spam update this year, and it comes with a longer window than usual: Google says it may take up to two weeks, while the August one finished in less than three days. As usual, Google didn’t say which tactics it’s going after.

That means the next couple of weeks will be full of people opening Search Console and asking “did we drop?” (I get some version of that message from clients after every update). It’s a fair question, but it’s too broad to tell you much. A site can gain clicks overall while its most fragile queries quietly slide to page two.

So before September’s results start coming in, we went back to the August 2026 spam update, which rolled out from August 18 to 21. We went through our clients’ Search Console data and asked a narrower question: did it matter where on page one a query was sitting when the update hit?

It mattered a lot. Queries near position 1 barely moved. Queries at the bottom of page one moved more often than not, and one in six of them ended up on page two.

(One note before the numbers: this is a before-and-after comparison. It shows what moved, and it can’t prove the update was the reason. I’ll come back to that.)

The short version

• The top barely moved. Near position 1, 11.1% of queries moved more than one spot. Between positions 7 and 10, 57.9% did.
• The bottom of page one is where you lose it. 16% of queries that started at positions 7–10 fell to page two. In the top 3, it was 0.4% (23 out of 5,881).
• New page-one arrivals came from page two. Of the 1,647 queries that climbed onto page one, 96.2% started at positions 11–20.
• Large sites gained clicks, small sites mostly didn’t. The median large site gained 8.2% and the median small site lost 6.6%, but the small sites have too little data to call that a pattern.

Who did we look at, and how?

We looked at Search Console data from a group of our client sites, split in two. Large sites had more than 100,000 US desktop clicks in the two weeks before the update. Small sites had under 5,000. They cover marketplaces, online learning, B2B software, ecommerce, logistics and management software.

We compared two 14-day windows: August 4–17 (before) and August 22 to September 4 (after). Both windows have the same mix of weekdays, and we left the rollout days out completely. We also pulled July 21 to August 3, to check whether the sites were already moving before the update (several were).

Everything is filtered to US, desktop, web search and final data. Desktop is one slice of US traffic, so read every number here as “US desktop” and nothing broader.

For the ranking analysis, the unit is a site–query pair: one query for one site that appeared in both windows with at least 50 impressions each time. That gave us 25,069 pairs.

What happened at the top of page one?

Very little. We had 2,641 queries with an average position of 1.5 or better before the update. Only 11.1% of them moved more than one position afterwards.

Three out of four (75.8%) were still at 1.5 or better two weeks later. Just 0.2% dropped out of the top 10, which is five queries out of more than 2,600.

One detail on wording. Search Console gives you your own site’s average position. So “still at 1.5 or better” means your average stayed there. It doesn’t tell us whether the same domains held positions 1, 2 and 3 in the actual results page.

My read is that a page holding position 1 has usually given Google plenty of signals to be confident about it. A spam update goes after low-quality and manipulative pages, and those rarely sit in the top spot for long. So the top of page one had less to lose.

What happened at the bottom of page one?

The further down page one a query started, the more it moved. At positions 7–10, 57.9% of queries moved more than one spot, about five times the rate near position 1.

And at the bottom of page one, moving often means moving to page two. Of the 4,854 queries that started at positions 7–10, 775 ended up with an average position past 10 (16%). Among queries that started in the top 3, that happened 23 times out of 5,881 (0.4%). So a query at the bottom of page one was about 40 times more likely to fall off it.

Part of this is just how rankings work. Around position 9, the pages next to you are usually close in quality, so a small change in how Google scores one of them is enough for two pages to swap places. At position 1, the gap to number 2 tends to be wider.

We didn’t run the same calculation for a two-week period without an update, so some of this movement is the normal churn you’d see in any fortnight. What we can say is that the gap between the top and the bottom held when we changed the minimum impressions to 20 or to 100 (details in the methodology).

Where did page one’s new arrivals come from?

Almost all of them came from page two. We counted 1,647 queries that went from an average position past 10 to 10 or better. Of those, 1,585 (96.2%) started between positions 11 and 20. Jumps from page three or deeper onto page one were rare: 62 in total.

There’s a catch in how this was measured. We only counted queries that showed up in both windows with at least 50 impressions. A page that was invisible before the update can’t appear in this list, so big jumps “from nowhere” are undercounted by design.

Even with that caveat, the practical takeaway is useful. If you’re wondering where your next page-one wins will come from after an update, your page-two queries are the list to start with.

Did large and small sites fare differently?

On clicks, yes. Every large site in the group gained clicks after the update, with a median of +8.2%. Most of the small sites lost clicks, with a median of −6.6%.

It’s tempting to read this as “the update favored big sites”. I wouldn’t, for two reasons.

• First, this is a group of sites only and not a full read of the whole world wide web. 
• Second, some sites flipped direction. One small ecommerce site was up 17.6% in the two weeks before the update and down 5.1% after it. One large site did the opposite: down 3.5% before, up 13% after. Without knowing what else changed on those sites, I can’t pin that on Google.

Also, every site’s average position improved, which sounds better than it is. Average position moves when the mix of queries changes, so it doesn’t mean every query got better.

So what should you do while the September update rolls out?

Mostly, wait until it’s done. The September rollout may take up to two weeks, and any data from inside that window mixes “before” and “after”. Once Google marks it complete, this is how I’d read the results, based on what we saw in August.

Look closely when:

• You have important queries sitting at positions 7–10. In August, those were about 40 times more likely to fall to page two than queries in the top 3. Pull them from the two weeks before September 24, so you have a clean comparison once the update finishes.
• A query falls from 7–10 to page two and stays there. One bad week is noise. Two full weeks after the rollout ends is a pattern worth working on (content depth, internal links from your strongest pages, a closer match to what people are searching for).
• You have a lot of queries at positions 11–20. That’s where 96% of August’s new page-one queries came from. If you’re choosing where to invest after an update, start there.

Don’t panic when:

• Your top-3 queries wobble during the rollout. In August, three out of four queries near position 1 ended up where they started.
• Total clicks move and you haven’t checked the trend before. Several of our sites were already up or down before August’s update began. Compare with the weeks before the rollout, too.
• Your windows don’t match. Compare the same number of days, with the same weekdays, and leave the rollout days out. Otherwise part of the change you see comes from the calendar (a window with more weekdays looks different from one with more weekends).

Where does this leave you for September?

After an update, the total clicks chart is the first thing everyone opens, and it tells you the least. Splitting your queries by where they started tells you much more about what happened and where to work next.

In our August data, the top of page one stayed where it was, the bottom moved a lot, and the next wave of page-one queries came from page two. We’ll run the same analysis once the September rollout is complete, and I’m curious whether it looks the same with a two-week rollout.

In the meantime, the most useful thing you can do costs nothing: export your queries at positions 7–20 for September 10–23, the two weeks before the rollout started. That’s your clean “before” to compare against.