Fact-Check: Nozzle's "Poisoned Google SERPs"

Every morning, rank tracking tools ask Google the same searches your customers type, then report where your site showed up. Since September, Google has been handing some of those tools a page that looks exactly like a normal search result and isn't. The usual businesses are missing, strangers are sitting in their spots, and nothing on the page says anything is wrong. If your report told you your site vanished from Google sometime in the last month, there is a fair chance your site never moved at all.

The short version, without the jargon

Think of a rank tracker as a mystery shopper. It walks into Google thousands of times a day, asks a question, writes down what it was shown, and hands you the report. That works as long as Google treats the mystery shopper like any other customer.

In September, Google started spotting the mystery shoppers. Some of them got turned away at the door. Others were let in and shown a fake display: the right shelves, the right signs, the wrong products. Real customers kept seeing the normal page the whole time.

That is the whole story, and it is why this is worth five minutes of a business owner's time. Plenty of people get a weekly ranking email from an agency or a tool. In late September a lot of those emails said something alarming. People have fired agencies, rewritten pages that were fine, and panicked about Google updates on the strength of numbers like these. If your sales, calls and form fills held steady while your ranking report fell off a cliff, believe the sales.

Local businesses should pay the closest attention. In the study we are checking, "near me" searches, small long-tail searches and local services such as law firms were hit the hardest. That is exactly where most small businesses compete.

What Nozzle found

Nozzle is a rank tracking company. On October 10, 2026 it published a study of 27.5 million search results pages it had collected for customers across 49 countries since September 13. [1] It called a page "poisoned" when fewer than half of a search's usual websites appeared on it, and those same websites came back on a later check. That second condition is the clever part. A real ranking loss sticks. A poisoned page goes away on its own.

By Nozzle's count, 8.0% of the pages it could judge were poisoned after mid-September, against 0.17% in the weeks before. The worst stretch was September 27 to 29, peaking at 27.0% in a single day. Most poisoned pages were one-offs: about four in five lasted a single check, and 77.9% had their normal results back within three days. [1]

Source: Nozzle, October 10, 2026. [1]

The swapped-in results leaned toward big platforms. Reddit, Facebook, LinkedIn, Quora and Instagram all showed up more often on poisoned pages, while YouTube, Yelp and MapQuest showed up less. One of Nozzle's examples was a US search for white-label local SEO services: nine relevant agencies on page one on October 4, all nine gone by October 7, replaced by a steakhouse, a public health association and a government site. [1]

The AI Overview rows deserve attention from anyone tracking AI visibility. Tools that report whether your business appears in Google's AI answers get that data the same way rank trackers do. If the page was poisoned, the AI Overview reading was too.

What holds up, and what needs a caveat

The event itself is well supported. Search Engine Roundtable reported heavy blocking of tracking tools on September 18, with Nozzle describing a drop of about 80% in the data it could collect and another rank tracker's status page saying Google had changed how it delivers results. [3] SEOmonitor says about 8% of the daily results Google returns to it have come back incomplete or incorrect since late September. [4] DemandSphere reports spurious results on roughly 11% of its daily checks. [5] Morningscore compared its tool against an ordinary browser at the same moment and found the browser got full results while the tool got cut-down pages. [6] Advanced Web Ranking was describing two different versions of Google, one for people and one for suspected bots, before Nozzle published anything. [7]

Those other figures sit near Nozzle's 8.0%, but each company counts something different, so read them as "same ballpark", not as confirmation of an exact number.

Four caveats belong next to the headline figure.

The 8% is a share of what Nozzle could judge. Of the 27.5 million pages collected, about 17.8 million had enough history to classify. Against the full set, the poisoned share works out to about 5.1%, a calculation Elsop made from Nozzle's own published files. [1][10]

It depends on how a tool connects to Google. Nozzle saw 1 to 7% poisoned through one network route and 9 to 30% through another over the same ten days. [1] Morningscore reported a similar gap between providers. [10] Your tool may have seen far more or far less than the average.

The timing overlaps a real Google update. Google's September 2026 spam update ran from September 24 to October 8. [9] Elsop found that 93.6% of Nozzle's poisoned pages fall in that window, which holds 68% of the pages Nozzle judged. [10] Nozzle's study does not mention the update. Its method does partly answer the objection, since a spam demotion that sticks would not be counted as poisoning, and Morningscore's side-by-side browser test points the same way. [6] Still, a site that dropped during those two weeks and stayed down may have a real problem.

Nobody knows why for certain. Nozzle's study states that Google is responding to AI tools scraping its results. Several people in the industry share that view, and Google has spent the past year making automated collection harder. [2][8] But Google has not commented on this, and the word "poisoning" is Nozzle's, not Google's. [8] Treat the motive as a reasonable guess.

What to check if your rankings dropped

None of this needs special software. It needs a few minutes and a little patience.

1. Open Google Search Console first. It reports what real searchers saw and clicked, straight from Google. If impressions and clicks held steady while your tracker showed a crash, your site did not drop. Nozzle gives the same advice. [1]
2. Wait three days before acting on any single drop. In Nozzle's data, most poisoned results cleared within that time. A real drop is still there next week.
3. Compare against the usual picture, not yesterday. If the sites around you on a search are strangers that have never ranked there before, such as a steakhouse on a search for marketing agencies, you are probably looking at a bad reading.
4. Ask your tool or agency one question. Do they detect and throw out bad readings from Google? Several tracking companies say they now do. [4][6] If the answer is vague, discount their September and October numbers.
5. Expect noisier AI Overview reports. Third-party tracking of Google's AI answers comes from the same collection, so it took the same hit. Weight your AI visibility reporting toward things you measure yourself: Search Console, referral traffic and leads.

If the phone kept ringing, the drop in your ranking report probably didn't happen.

One last point, because the name causes confusion. "Poisoned search results" is also the term for scam and malware sites pushed to the top of Google by attackers. That problem is real and ongoing, but it is a different one. [11] Nozzle's study is about Google showing bad data to tools, not about bad sites being shown to people.

Nozzle's full study, including its method and the raw data files, is worth reading in the original: Poisoned Google SERPs. [1]