Google may be sending rank trackers fake search results

Google appears to have started serving deliberately altered search results pages to automated scrapers since 13 September 2026, according to research from rank-tracking firm Nozzle. Roughly 8% of the 27.5 million search pages it collected came back "poisoned", meaning most of the usual ranking sites had been swapped out for others. Because many AI visibility and citation-tracking tools rely on the same scraping methods, this also distorts what brands think they know about their presence in AI Overviews.

What changed?

Nozzle published its findings on 10 October after analysing 27.5 million search results pages collected for customers in 49 countries between 20 August and 9 October 2026. It defines a page as "poisoned" when fewer than half of a keyword's usual top 10 sites appear anywhere on it, and those sites then reappear on a later fetch.

Before mid-September, this barely happened. Since 13 September, 8.0% of the pages Nozzle collected were poisoned, up from a baseline of just 0.17%, Nozzle reports on its research page.

The timing lines up with other changes at Google. The worst days were 27 to 29 September, when about a quarter of all results pages were poisoned, overlapping with Google's September spam update, according to Mean CEO's analysis. Requests to Google also started failing more often around 13 September, and the pages that still came through stayed normal until 20 September, when poisoned SERPs first appeared, according to TrueRanker.

Nozzle's theory is that this is a defensive move against AI scraping rather than an attack on rank trackers specifically. A refused request tells a scraper to try again, while a poisoned one gets stored as real, which makes it a more effective deterrent. Search Engine Journal reports the same reasoning, suggesting the aim is to thwart AI companies whose large language models lean on Google's results for grounding, while also confounding rank trackers along the way.

According to PPC Land's summary of related reporting, Google appears to be using three tactics together: blocking ranges of residential proxy addresses, showing far more captchas, and serving altered results to requests it suspects are automated.

Why does this matter for AI search visibility?

Poisoned pages do not just shuffle organic rankings. They also change what AI Overviews show and cite. AI Overviews appear on 34.9% of poisoned SERPs, down from 54.2% on normal ones, and the sources they cite overlap far less with the previous version, according to Brainz Digital's write-up of the Nozzle data.

Nozzle's own figures show the same pattern. On poisoned pages, the share of AI Overview sources carried over from the previous SERP fell from 68.1% at baseline to 37.5%. In plain terms, a tool checking a page before and after poisoning hit would see a very different set of AI Overview citations, even though nothing about the brand's actual content or ranking had changed.

The "People Also Ask" box is affected too, keeping only 46.9% of its questions compared with 87.4% on unaffected pages, per Brainz Digital's analysis.

This creates a direct problem for anyone measuring generative engine optimisation (GEO) or answer engine optimisation (AEO) performance. Any tool tracking whether your brand is cited in AI Overviews, or monitoring share of voice across AI search, is likely built on the same kind of automated scraping that triggers poisoning. If the underlying scrape is unreliable, every metric built on top of it is too.

The impact also varies sharply by country. Austria hit 18.8% poisoning, and Austria, Belgium and Switzerland were hit far harder than Germany, France or the UK, according to Mean CEO's analysis. A brand tracking visibility across multiple markets could see wildly inconsistent data depending on where its audience, and therefore its scraping requests, are based.

What should you do now?

Treat any sudden ranking or AI citation crash with suspicion before acting on it. On a poisoned page, the tracked site falls out of the top 10 58.6% of the time, so a rank tracker can show a crash that never actually happened, Mean CEO explains. Cross-check any alarming drop against Google Search Console, which reflects real impressions and clicks rather than a scraped snapshot.

This is also not necessarily permanent. Of the poisoned SERPs recorded since 13 September, 67.9% had their usual sites back within three days, per Nozzle's data. A dip that corrects itself within a few days is more likely to be poisoning than a genuine algorithmic demotion.

Be sceptical of any single-source AI visibility report during this period, particularly if it relies on scraped SERPs rather than direct platform data or real conversation logs. Several rank-tracking vendors have already flagged the issue to customers; one confirmed it had been seeing the poisoned SERPs problem since late summer and put additional checks in place to filter and double-check the results it returns, as reported by PPC Land.

Given how much noise is now baked into scraped SERP data, it is worth checking your own AI visibility through a method that does not depend on guessing what Google shows a bot. This is one reason tools like Sited's free audit at https://sited.online focus on how your brand actually appears across AI answers, rather than reverse-engineering a single poisoned snapshot.

Finally, if you manage multi-market campaigns, weight your trust in the data by country. Markets like Austria, Belgium and Switzerland are seeing far higher poisoning rates than the UK, France or Germany, so treat visibility swings in those regions with extra caution until the picture settles.

Frequently asked questions

Is Google confirming it is poisoning search results?

No. Google hasn't commented, according to Mean CEO's reporting, so the explanation that this targets AI scrapers and grounding remains Nozzle's interpretation of the pattern rather than an official statement.

Does this affect my actual Google ranking, or just what tools report?

Based on the research, it appears to affect what automated tools see, not necessarily your real position for genuine human searchers. Checking Search Console or doing a manual, logged-out search is the safer way to confirm a real change.

How can I tell if a ranking drop is real or a poisoned SERP?

Look for a quick bounce-back. Since 67.9% of poisoned SERPs had their usual sites restored within three days, a drop that reverses quickly without any changes on your end is a strong signal it was poisoning rather than an algorithm update.

Does this affect AI Overview citation tracking specifically?

Yes. AI Overviews show up on 34.9% of poisoned SERPs, down from 54.2%, and the sources they cite overlap far less with the previous version, so any tool measuring your AI Overview citation rate from scraped data could show a distorted picture during this period.