Google appears to be serving altered results pages to rank-tracking tools, according to a Nozzle report published on October 10, 2026. Since October 1, between 4% and 11% of the searches it records each day have come back with different results, after a peak above 25% in late September. This matters to anyone who measures their SEO with these tools.
What we know
- Who says so: Nozzle Research, the analysis team at Nozzle, a company that sells rank tracking. Its report, "Poisoned Google SERPs," is dated October 10, with data through October 9. Google has not confirmed anything.
- What "poisoned" means: a complete results page, with no error or captcha, missing more than half of the sites that usually appear for that query, which then return on a later search.
- Sample size: 27.5 million results pages collected since September 13 in 49 countries, on desktop, iPhone and Android, out of a 63.1 million sample going back to August 20, according to Nozzle. The "before" figures come from its baseline period, August 20 to September 9.
- When it started: per the report, requests to Google began failing on September 13 and altered pages appeared on September 20. They were 0.17% before and 8.0% since September 13. They peaked at around 25–27% on September 27–29 and have run at roughly 4–11% a day since October 1.
- How much changes: an altered page keeps only 20.7% of the previous query's sites, versus 88.5% in the baseline.
- How long it lasts: 80% of episodes last a single query, and in 77.9% of cases the usual sites return within three days.
- Telltale signs: fewer AI Overviews (34.9% vs 54.2%), fewer "People also search for" blocks (52.7% vs 87.2%), more image packs (44.2% vs 23.0%), and more Reddit and Facebook.
- The effect on reports: a project's own site dropped out of the top 10 on 58.6% of altered pages, versus 2.7% in the baseline.
- Nozzle's explanation: it sees this as Google's response to scraping for AI training and grounding, which hits rank trackers because they collect results the same way. That is its interpretation. Search Engine Journal's author adds, as his own opinion, that it may also target spammers.
What changes and what doesn't
The report covers the pages that automated tools receive. It does not measure what people see, and it does not say your real rankings change.
What can change is what your tool shows you: a one-day drop that never happened. Nozzle lists limits: the latest days are still filling in, and the rate varies by request route (1–7% on one, 9–30% on another, on the same days). It is a company in the rank-tracking business measuring in its own way; it is not Google data.
How to tell if it affects you
- Check Search Console. It is Google's own data on your site. If a page "disappears" in your tool but still has impressions and clicks in Search Console, nothing happened to it.
- Compare with weeks, not yesterday. Yesterday may be altered too. Look at which sites usually rank for that query over recent weeks.
- Wait for it to repeat. A real drop persists. If everything is back after two or three days, it was noise.
- Ask your provider. If you pay for a rank tracker, ask whether it detects and discards these pages, and since when.
- Don't change a page over one data point. Rewriting titles or content after a one-day drop can break what was working.
Related: Search Console Now Lets You Filter by Several Countries
Sources
Updates: if Google comments or Nozzle updates its data, it will be added here with a link.