Rankings Were Volatile From September 4 to 13 and Google Announced Nothing - Seen From Two Sites

In short: on September 4 trackers recorded a sharp spike in Google ranking volatility, results stayed unstable for nine days, and by September 13 they had settled back. Affected sites recovered. Google announced nothing, neither at the start nor at the return. These dates appear nowhere in the official update history, so calling it an update rests entirely on tracker readings.

I have two projects of very different weight on hand, so I checked both. The result is more interesting than I expected: on the large one the update is absent where you would look for it, and present where you would not.

What happened

On September 4, AccuRanker, Mozcast, Semrush and the rest lit up at once. Large content projects reported visibility losses. Google said nothing.

By September 13 the volatility had subsided. Positions came back, Discover normalised over the weekend. The industry described this as a rollback, but again without confirmation from Google. The timeline is documented at Search Engine Roundtable; everything else I found retells it.

Nine days from spike to recovery is a short cycle. A core update usually takes two or three weeks and gets finished even when the market complains. Here everything wound down halfway, which itself argues against the core update theory.

The project big enough to measure on

A cleaning company in St Petersburg whose marketing I run. Over 28 days: 1,437 clicks and 127,844 impressions, average position 11.3. At that volume you see structure rather than noise.

Daily numbers across the update window:

Now the part that matters. In the thirteen days before the update (Aug 20 - Sep 1) the site averaged 50.1 clicks a day. In the thirteen days covering the spike and the recovery (Sep 2-14) it averaged 60.5.

Traffic did not drop. It grew by a fifth.

And here I have to be honest rather than clever: I cannot attribute that growth to the update. September is the start of the season in cleaning - offices and schools come back, demand rises on its own every year. The rise began on September 7 and is still holding. Separating season from algorithm on this data is impossible, and anyone who tells you otherwise is selling you a report, not an analysis.

What can be stated firmly: on a project where Google shows the site 127,000 times a month, a nine-day update left no trace in clicks that you could see with the naked eye. No drop, and no bounce after September 13 either.

September 11 stands out. On both sites at once

Look at the Sep 11 row above: position 15.8 against a usual 9-11.

Now the second project - my own site, two orders of magnitude smaller: 81 clicks and 7,813 impressions over the same 28 days.

September 11, both sites. Different niches, different volumes, different semantics - and the same average-position spike on the same day, back to normal the next morning.

One site is an anecdote. Two unrelated sites with the same anomaly on the same date is an observation.

Why this is arithmetic, not a drop

I broke September 11 down by page on both projects, and on neither did the core move.

On the cleaning site the workhorse is a post about cleaning after pest control: 748 impressions at position 4.4 and 22 clicks. The next day, September 12: 645 impressions at 4.1 and 25 clicks. The page on office sanitary standards: 293 impressions at 5.3. Exactly the values they show on any other day.

On my own site the same: the main post held at 7.3, its English version at 6.2, while the average was dragged down by two deep-position visitors - the services page took 67 impressions at position 68.8 on a query GSC will not even show me, and another post picked up 11 impressions at position 51.

The mechanism is identical. When results are shaken, pages get pulled into queries they have no business appearing for, at positions nobody scrolls to. More impressions, worse average, unchanged real positions. On the summary line it reads as a collapse; broken down by page, nothing happened.

First practical conclusion: during volatility, average position is a useless number. Watch the pages that bring clicks, and the specific queries.

Second conclusion: scale decides

When I first looked at this update using only my small site, the picture came out quite different - September 11 looked like an event and everything else looked like noise you cannot extract anything from.

On the large project it turned out there is exactly one event, and it is not about losing traffic.

If your site gets a few hundred impressions a day, the honest answer is that your own data will tell you nothing about Google updates. Any movement drowns in variance. Look either at a project with enough volume, or at volatility trackers that aggregate thousands of sites.

What to do

Nothing. That is the answer.

Unannounced volatility is not an assessment of your site. Every change you make in panic on day nine of the shaking is a change you will be undoing later.

The one thing worth doing in advance is recording the numbers - daily, before, during and after. Then the next time traffic drops you can tell an algorithm update apart from your own broken robots.txt, and those are two very different conversations with a client.

I keep those snapshots every few days on both projects. That is the only reason I have two tables here instead of a retelling of somebody else’s volatility chart. It takes half an hour to set up and pays for itself on the first unexplained drop.

News source: Search Engine Roundtable, Google Search Ranking Update Around September 4th That Reverted September 13th - the timeline and a summary of what the trackers showed.

Data source: Google Search Console for both projects, exported on 17 September 2026. GSC data arrives with a two to three day lag, so the last complete day in the tables is 14 September.