On October 10, the rank tracking company Nozzle published research on what it calls poisoned Google SERPs. Across 27.5 million search results pages in 49 countries, 8.0% came back wrong between September 13 and October 9. The baseline was 0.17%.
Not missing. Not erroring. Wrong.
Two days later, Search Engine Roundtable confirmed the problem is industry wide, pointing to notices from SE Ranking, DataForSEO, Sistrix, Advanced Web Ranking and Morningscore.
The obvious reaction is to file this under vendor problems. Your tool hiccuped, Google will patch it, switch platforms if yours was slow.
That read is comfortable and wrong. The tool worked exactly as designed. It asked Google a question, Google answered, and the answer was fabricated.
TL;DR
- Nozzle found 8.0% of 27.5 million tracked search pages returned swapped-out results between September 13 and October 9, against a 0.17% baseline. The daily rate hit 27.0% on September 29.
- On a poisoned page, the tracked site was missing from the top 10 in 58.6% of cases. The normal rate is 2.7%. The cliff on your September chart may not have happened.
- DemandSphere independently put it near 11% of its daily SERPs. Morningscore saw about 60% of its searches truncated at the September 30 peak.
- The worst days sat inside Google's September 2026 spam update window, so every phantom drop came with a ready-made explanation.
- Google has not commented. The mechanism points at AI scrapers, not at you, which is why waiting this out is not a plan.
What Poisoned Google SERPs Look Like From the Inside
Nozzle's description is the part worth memorizing: "A refused request tells a scraper to try again, while a poisoned one gets stored as real."
Google returns a full page. No error code, no captcha, no rate limit warning. Same template, plausible titles and URLs. The only thing wrong is which sites are on it. Nozzle measured 20.7% of a keyword's usual organic sites surviving onto a poisoned page, against 88.5% normally.
A tool that gets an error can flag the error. A tool that gets a clean page stores it as fact.
That is how three weeks of fiction got written into thousands of ranking histories before anybody caught it.
It is not one company's bad week either. DemandSphere's Ray Grieselhuber wrote on October 7 that Google has "started to introduce spurious results into their SERPs when they detect alleged bot traffic," affecting about 11% of his company's daily SERPs. Morningscore's Karsten Madsen published a timeline on October 8 showing about 1,800 tracked websites more than 25% below their September 23 score at the worst point.
Three companies, three methods, one conclusion. All three sell SERP data, which is worth saying out loud. It is also why they noticed.
The Timing Is the Part That Will Cost People Money
Google's September 2026 spam update started rolling out September 24 and finished October 8. Morningscore's truncated results started the evening of September 24. Nozzle's poisoning peaked September 27 through 29. Both windows sit inside the spam update.
So for three weeks, any owner watching rankings fall had a confirmed Google update to blame, a plausible story, and a chart that agreed. A two-week rollout already makes a drop hard to date. This is worse than undatable. Some of those drops were never real.
The signature is a collapse across many keywords at once, followed by recovery inside three days. Nozzle found 77.9% of poisoned pages back to normal within three days. Morningscore found 74% of missing sites back in the top 20 within three days.
Local and Legal Took the Worst of It
Nozzle broke the data down by query type. "Near me" queries were poisoned at 11.4%, against 7.8% for everything else. The most navigational, brand-style queries were cleanest at 5.3%. By US industry, legal was highest at 15.9%, with home services next at 12.7%.
Read that again if you run a home services company in Boca Raton or a practice in West Palm. The queries that make your phone ring are the queries where the tracking data is least trustworthy right now.
You Are Not the Target, Which Is Why It Will Not Stop
Here is what the "switch vendors" advice misses.
Nozzle calls it Google's response to AI scraping and grounding, the bulk collection of search results to feed AI answers. Grieselhuber put it bluntly on September 24: "Google does not want the AI labs to use their SERPs as part of their live retrieval process to answer questions." Barry Schwartz at Search Engine Roundtable reached the same read.
Rank trackers collect pages the same way AI retrieval systems do, and Google's spam policies have long listed "scraping results for rank-checking purposes" as prohibited machine-generated traffic. So the tools were never welcome. They are now standing in the blast radius of a much bigger fight.
Google has not commented, so treat the motive as inference. A statement from Google or a sustained return to baseline would settle it. Neither has happened. Nozzle's daily rate was still running at about 5 to 11% in early October.
The Real Cost Was Never the Report
Nobody loses money because a dashboard number was wrong. They lose money because of what they did next.
Rewrote a page that was ranking fine. Killed a content line that was working. Cut spend on the one channel producing. Fired an agency over a September chart.
That bill comes due in October and November, not September.
What To Do This Week
- Quarantine September 13 through October 9 in your ranking data. Do not delete it, label it. Then pull the same range in Google Search Console, which reports impressions and average position from Google's own data and was never scraped. If Search Console held steady while your tracker shows a collapse, your rankings did not move.
- Ask your SEO vendor two questions in writing. Did your rank data between September 13 and October 9 come from scraped SERPs, and did you detect, discard or backfill poisoned results? Morningscore publicly documented catching it and correcting its records. Keep whatever answer you get. If you want a second read on the answer, that is part of our SEO services.
- Reverse any decision from the last three weeks that rested only on a ranking chart. Pages depublished, budget moved, content paused, vendors put on notice. Re-check each against Search Console impressions, form fills and booked calls.
FAQ
Did my Google rankings actually drop in September 2026?
Possibly not. Three vendors reported Google returning falsified or truncated results to their crawlers between mid-September and early October, with Nozzle measuring 8.0% of pages affected and the tracked site missing from the top 10 on 58.6% of those. The reliable check is Google Search Console, which uses Google's own data rather than scraped pages. If impressions and average position held steady there, your rankings held steady.
Is Google blocking SEO tools on purpose?
Google has not commented publicly. Its spam policies have long prohibited scraping results for rank checking, but the vendors investigating believe the new measures target AI systems scraping live search results, with rank trackers caught in the same net because they collect pages the same way. That is an inference from observed behavior, not confirmed Google policy.
How should I measure SEO performance if rank data is unreliable?
Use data you own. Search Console impressions and clicks, organic sessions by landing page, form submissions and calls by source, cost per booked job. Rank position was always a proxy for those things, which is the case for analytics that drive smarter marketing decisions. This month it became a proxy you cannot audit.
The Reframe
For years the industry treated a scraped ranking number as an observation. It never was one. It was a reconstruction, assembled by a robot pretending to be a customer, from a company that has now decided it does not want robots pretending to be customers.
The number did not break in September. It got caught.
Note: figures in this post come from Nozzle's original October 10 edition, with data through October 9. Nozzle has since extended the study and its live page now shows later figures.
Book a 30-minute strategy call. No pitch, just a straight read on whether your September reporting told you the truth and what to do about the decisions you already made off it. Get a free strategy call.