Google traffic that dropped in September often comes from three breaks outside your own site: product carousels gone from the EEA, Shopify's bot filter and the spam update, after which the knowledge base of Oase Creative recovered without intervention.
Why did your Google traffic drop in September?
Google traffic dropped in September? Chances are it wasn't you.
Within four weeks, three things happened that each put a kink in your numbers, and none of them started on your own site. Google removed organic product carousels in Europe. Shopify changed how it counts sessions. And Google ran a spam update that took almost two weeks. Anyone who puts Q3 and Q4 side by side without those three dates will draw the wrong conclusion, and may start rewriting things that aren't broken.
| When | What happened | What you see | Who it affects |
|---|---|---|---|
| 15-16 September | Organic product carousels disappear from Google in the EEA | Fewer impressions and clicks on product searches | Online stores with product pages |
| 21-23 September | Shopify filters bots out of session reports by default | Fewer sessions, higher conversion rate, same orders | Every Shopify store |
| 24 September - 8 October | September 2026 spam update | Positions that drop and then come back | Everyone who ranks in Google |
Two of these are real: the carousels cost visibility and the spam update shook up positions. One is only a different yardstick. That difference decides what you do.
We saw it in our own data too. Seven keywords from our own knowledge base fell during the spam update from the top 5 on average to around position 10, and were back close to where they were in the week around completion. We rewrote nothing in that time. Below, per break: what happened, which annotations to add, how to compare fairly, and when it is your own site after all.
What happened to product carousels on 15 and 16 September?
The blocks of popular products at the top of Google have all but disappeared in the EEA. That is not an algorithm update but the result of a decision by the European Commission.
On 23 July 2026 the Commission found that Google favours its own services in Search. It named shopping, hotels, transport and sports results. Google received a fine of €460 million under the Digital Markets Act. According to the Commission's press release, Google had sixty days to comply, or face periodic penalty payments.
On 8 September Google published a page on regional differences in the search experience. It describes three new elements for the EEA: a block for comparison sites, a block for direct suppliers, and carousels based on structured data. A week later Productrise saw product carousels almost completely disappear within a few days in Germany, France, Belgium, the Netherlands and Sweden. In the United Kingdom, the United States and Australia nothing moved. Productrise sells tracking for those carousels itself, so read it as a vendor measurement.
What this means for your numbers: if your store got organic traffic through those carousels, that traffic has been structurally lower since mid-September. That is a real loss of visibility, not a measurement error, and it is not caused by your own SEO. The Shopping tab is gone from the menu, but still works through a parameter in the search URL.
It isn't over yet. The Commission is still assessing Google's changes to Shopping Ads, and the conversation about AI Overviews and AI Mode is still ongoing. How to get your products visible for free in the new layout is covered in detail in our piece on the disappearing product carousels. In short: markup on your category pages matters, and that is something you control.
Why do you see fewer Shopify sessions since 21 September?
Because Shopify started counting differently. Your visitors didn't leave.
Between 21 and 23 September Shopify changed its session measurement. Shopify's Help Center explanation says it plainly: identified bot sessions are filtered out of session-related reports by default. On top of that, sessions no longer end at midnight UTC, and some sessions without a page view now count. Shopify itself advises using data from after the update as your new baseline.
Usually it plays out like this: your orders stay the same, your sessions drop, so your conversion rate goes up. Anyone who sees that in a monthly report thinks the new product page works or the campaign landed. Your customers didn't change, your yardstick did. Shopify also mentions the opposite outcome: if more sessions without a page view now count in your store, your conversion rate can drop. So look at your own report.
How many bots disappear differs per store. Shopify calls the filter deliberately conservative and does not show what it removes. The only measurement we found comes from Harvv, a vendor of bot measurement, using its own pixel on Shopify stores between 29 September and 7 October. For the median store in that sample a small share of sessions was bot traffic, but the spread was wide: the busiest store was many times higher. Almost none of those bots reached the cart. AI assistants were only a fraction; most of it was ordinary automation posing as a browser.
One vendor with its own product, one measurement period, no second source: that is the limitation. The takeaway: an average across your whole store can hide a week in which a large share of your sessions were not human. So don't look at the average, look at your own sessions per week around 21 to 23 September.
What did the September 2026 spam update do to our own knowledge base?
It pushed it down and brought it back, without us doing anything. We measured that from start to finish.
According to the Google Search status page, the update started on 24 September at 16:15 UTC and was complete on 8 October at 08:00 UTC. That is 13 days and almost 16 hours, more than five times as long as the August spam update. Search Engine Roundtable saw the biggest swings around 25 to 27 September, on 30 September and from 4 to 6 October.
We track seven keywords for which articles from our knowledge base rank. From Search Console, in three one-week windows (Dutch keywords, Dutch decimal commas):
| Window | Weighted average position | Keywords in the top 10 |
|---|---|---|
| 18-24 September (before the update) | 4,5 | 7 of 7 |
| 27 September - 3 October (during) | 10,6 | 2 of 7 |
| 4-10 October (around completion) | 5,1 | 7 of 7 |
Two keywords showed the pattern most clearly:
| Keyword | Before | During | After |
|---|---|---|---|
| hoger ranken op bol.com (rank higher on bol.com) | 2,1 | 10,4 | 3,1 |
| bol.com seller tips | 5,7 | 18,4 | 4,7 |
The cleanest evidence is our page on ranking higher on bol.com. We hadn't touched it since 4 September, and it recovered anyway. So nothing on the page explains the recovery. What changed was Google.
Our service pages held up reasonably well in those weeks; the knowledge base moved the most. That is one measurement on one site with seven keywords, so not a rule for everyone. But it is a fully measured pattern: dropping during the update, recovering in the week around completion.
Why did we change nothing during the spam update?
Because we had written down in advance when we would step in, and that line was never crossed.
In the week to 3 October only two of the seven keywords were still in the top 10. The temptation at that point is to overhaul titles, rewrite paragraphs or start shuffling internal links. We didn't. On 6 October we set one decision rule: step in if a page still has an average position worse than 8 after 10 October. After 10 October that applied to none of the seven.
There is a reason to wait. During an update you don't know which of your changes does anything. Rewrite on day five and recover on day fifteen, and you'll think your rewrite worked. You take that lesson into the next update, and then it doesn't hold. You also lose a clean measurement, because you no longer know what caused the recovery: Google or you.
There is a practical reason too. A page that has ranked well for weeks and then drops in a window where everyone is swinging probably doesn't have a problem you caused. Start tinkering then, and you risk removing exactly the thing that made the page rank.
Waiting is not a tactic that always works. It works when your pages were stable before the update, when the drop lines up with the update window, and when you have a line at which you do act. Without that line, waiting is just putting it off.
Which annotations should you add in GA4, Search Console and Shopify?
Three dates, in every report you look at. Then three months from now you will still see why the line kinks.
An annotation is a short note on a date in your reporting. In GA4 you can add it directly to the charts. For Search Console and Shopify reports, note it in your own reporting or in the document where you keep the exports. The tool doesn't matter. What matters is that everyone who looks at the chart later sees the break.
| Date | Annotation | Add it to |
|---|---|---|
| 15-09 | Product carousels gone from Google in the EEA | GA4, Search Console, your reporting |
| 21-09 to 23-09 | Shopify filters bots out of sessions, new baseline | Shopify reports, GA4, your reporting |
| 24-09 to 08-10 | September 2026 spam update | Search Console, your reporting |
| 25-09 to 27-09, 30-09, 04-10 to 06-10 | Peaks in volatility during the update | Search Console (optional) |
Mind the difference between GA4 and Shopify. The Shopify change is about the reports inside Shopify, not GA4. If sessions in the two have drifted further apart since late September, this may be the cause. Don't assume it blindly: compare per week what each of them counts. How to make traffic from AI assistants visible separately in GA4 is in measuring AI referrals in GA4.
If an agency or freelancer builds your reporting, ask whether these three dates are in it. If not, you'll get a story in the quarterly review about sessions going down and conversion going up, and advice built on a measurement error.
How do you compare Q3 and Q4 fairly?
On orders and revenue, not on sessions and conversion rate. And for rankings, with windows the spam update falls outside of.
None of the three breaks changed the yardstick for orders and revenue. They can really drop, for example because the product carousels cost visibility. But a drop in orders is a real drop, and you can act on it. Sessions and conversion rate in Shopify since 21 September are no longer comparable with the months before.
| You want to know | Compare | Don't |
|---|---|---|
| Whether your store is doing worse | Orders and revenue, Q3 against Q4 | Sessions or conversion rate across 21 September |
| What the spam update did | Positions 10-09 to 23-09 against 09-10 to 22-10 | Positions from during the update |
| What the carousels cost | Clicks on product searches before and after 15-09 in Search Console | Total organic traffic without distinction |
| Whether your conversion really improved | Weeks from 23-09 against each other | Conversion in October against August |
The window 10 to 23 September against 9 to 22 October is clean for the spam update: two weeks before, two weeks after a day of rest. For the other two breaks that doesn't hold. The first window falls partly after the carousels and runs across the Shopify change. So use it for positions in Search Console, not for sessions in Shopify.
The second half of that window runs to 22 October. If you run this comparison earlier, your second window is shorter. That's fine, as long as you say so.
When is an SEO traffic drop your own site after all?
When the dates don't fit, when your pages don't come back, or when your orders fall too. Then it's time to look at your own site.
Three breaks in one month make it easy to blame everything on them. That is just as risky as rewriting everything. These signals point to something else:
- The drop doesn't match the dates. Did it start in August, or only in mid-October? Then the cause lies elsewhere.
- Your pages don't recover after 8 October. For us, all seven keywords were back in the top 10 in the week around completion. If a page is still well below its old level two weeks after the update, Google probably found something on it.
- Only one type of page drops. Are mainly your category texts or your blog dropping, while your product pages hold? Then look at what those pages have in common.
- Your orders fall too. A yardstick doesn't change revenue.
In those cases, first check Manual actions in Search Console. If something is listed there, a person at Google has reviewed your site and you know exactly what it's about. If nothing is there, read Google's spam policy on scaled content. It covers producing many pages mainly to rank in search results, with little original value, however that text was made.
If you had a lot of category or blog text generated with AI over the past year and those pages dropped during the update, that's the first place to look. What AI text does and doesn't do for rankings in e-commerce is in AI content and your rankings.
What do we do with this ourselves at Oase?
We put the three dates in every report, and we don't rewrite anything based on a month that went through three breaks.
For our own knowledge base that means: no changes to the seven pages that dropped and recovered. We'll run the comparison of 10 to 23 September against 9 to 22 October as soon as the second window is complete. If a page comes out of that which hasn't come back, we look at that page, not the whole knowledge base.
For online stores, in an SEO project we look at the dates first and the site second. For Shopify stores that also means taking conversion from 23 September as the new baseline, and answering whether a campaign worked with orders, not sessions. The carousel break is also a reason to check the markup on your category pages. That is the part of the new European product results you can influence yourself. If you build on Shopify, we help with that too through Shopify stores.
Seeing a drop that doesn't fit this list, or not sure which of the three hit you? Tell us where it's stuck. Send your Search Console export along and we'll look together at which date belongs to your kink.
