Google is poisoning search results it serves to rank trackers. If your tracker showed a page falling out of the top 10 in late September 2026 and coming straight back a day or two later, your site probably never moved. New research from Nozzle shows Google has been serving deliberately altered search results to automated SERP collectors, and rank trackers are caught in it.
Nozzle calls these poisoned SERPs: complete, normal-looking results pages where most of the organic results have been quietly swapped out. There is no captcha, no error, and no template change. The only giveaway is that the result set no longer matches the keyword's own ranking history.
This article breaks down the study's 27.5 million SERPs, explains what changes on a poisoned page, and shows how to separate a data problem from a real ranking drop before it ends up in a client report.
What a poisoned SERP looks like
Nozzle's clearest example is the query "white label local seo" on US desktop. On October 4 it returned the white-label SEO vendors you would expect. On October 7, the same keyword, device, and location returned a completely different page one.
Normal, Oct 4
- seohive.co
- synup.com
- brightlocal.com
- keyword.com
- merchynt.com
- seoreseller.com
- almcorp.com
- databloo.com
- quora.com
Poisoned, Oct 7
- localwp.com
- localjh.com
- reddit.com
- naccho.org
- usa.gov
- uncdf.org
- wa.gov
- localmontecito.com
- ohio.gov
A rank report reads that page as a total ranking collapse for every vendor on the original list. Nothing in the HTML says otherwise: the page uses the same template and the same frontend build as a normal Google results page.
When Google started poisoning search results
The sequence is what makes this poisoning rather than blocking. A refused request tells a scraper to retry. A poisoned response gets stored as real data and silently corrupts the dataset. Nozzle concludes this is Google's response to AI scraping and grounding, the bulk collection of search results to feed AI answers. Rank trackers collect SERPs the same way, so they get swept up.
Observation vs explanation. The data documents a sharp, measurable change in collected SERPs. The cause is a separate question. The timing fits Nozzle's AI scraping explanation, but timing alone does not prove intent, and Google has not confirmed how or why results were altered.
Writing for Search Engine Journal, Roger Montti suggested a further motive: the tactic may also frustrate spammers trying to reverse engineer ranking signals or distill AI Overview answers.
What replaces the usual results
When the usual sites disappear, social platforms and forums fill the gap. The chart below compares each keyword's poisoned SERPs with its own other SERPs over the same period.
Three or more social sites share the top 10 on 7.2% of poisoned SERPs, against 2.7% otherwise. On US keywords, foreign country-code domains such as .ca or .co.uk appear in the top 10 of 14.2% of poisoned SERPs, more than double the normal 6.1%.
SERP features change too, except the paid ones
Poisoning reaches beyond the ten blue links. For each part of the page, Nozzle measured how many of the previous SERP's sites, questions, or searches reappear.
What appears on the page shifts as well:
| SERP feature | Shown on normal SERPs | Shown on poisoned SERPs |
|---|---|---|
| AI Overview | 54.2% | 34.9% |
| People also search for | 87.2% | 52.7% |
| Image pack | 23.0% | 44.2% |
| Discussions and forums | 20.5% | 27.0% |
| Desktop pagination stopping at page 4 or earlier* | 0.2% | 50.9% |
*On keywords whose results normally run to page 8 or beyond.
The paid parts hold. Shopping results keep 92.2% of their items against 92.8% normally, and ads keep more than usual: 83.7% against 70.6%. The parts of the page advertisers pay for are largely left alone.
Where it hit hardest
By country
Across 49 countries, Western Europe ran lowest. The United Kingdom was lowest of all at 3.0%, with Spain, France, Italy, the Netherlands, and Germany all under 5%. Their neighbors were not: Austria hit 18.8%, Belgium 13.2%, and Switzerland 12.8%. South Africa (18.4%), the UAE (14.4%), New Zealand (13.8%), Norway (13.1%), and Saudi Arabia (12.9%) also topped 12%. The United States sat at 8.5%, with Australia (11.5%), India (10.9%), and Canada (10.5%) higher.
Device made little difference: 7.7% on desktop, 8.3% on iPhone, 7.1% on Android. The network route did matter. Over the same days, one of Nozzle's collection routes returned 1 to 7% poisoned SERPs while another returned 9 to 30%.
By industry
Legal SERPs were poisoned most, reaching 51.3% on September 27. Home services, a core vertical for local SEO, peaked at 48.4%. If you manage legal, home services, or transportation clients, late September data deserves a second look.
By search type
- "Near me" searches: 11.4% poisoned, against 7.8% for everything else.
- Long-tail keywords: 9.4% for phrases searched fewer than 100 times a month, 6.1 to 7.1% above that.
- Branded, navigational searches: lowest at 5.3%.
- Local targeting: no real effect. The same phrase tracked nationally and at city level, on the same device and in the same country, came out at 7.5% and 7.4% across 21,508 pairs.
How long poisoning lasts
Most poisoning is brief. 80% of spells affect a single SERP, 15% last two, and 5% last three or more. The next SERP has the usual sites back 55.9% of the time, and 77.9% recover within three days.
One subtler finding matters for reporting: even SERPs that were not flagged as poisoned got noisier after September 13. On those, organic results kept 75.5% of the previous SERP's sites, down from the 88.5% baseline, and tracked sites missed the top 10 on 10.2% of them, against 2.7% before. Volatility rose across the board, not only on the flagged pages.
How poisoned SERPs distort rank tracking
A rank tracker collects a SERP, finds the tracked domain, records its position, and compares it with earlier observations. That process assumes the returned page represents the real results. When it does not, four reporting problems follow.
| Distortion | What happens |
|---|---|
| False ranking declines | A usual top 10 site vanishes from one SERP and the tracker logs a loss that never happened for real searchers. |
| Misleading competitor comparisons | When a competitor drops out of a poisoned SERP, another domain appears to gain. Teams may credit a rival's content or links for a "win" that was only bad data. |
| Corrupted historical reporting | The bad observation stays in the database after results return, pulling down daily averages, visibility scores, and share of voice charts. |
| Wrong responses | Teams audit canonicals, redirects, and recent edits, or even roll back changes, to fix a drop that was never real. |
Why standard monitoring misses it
SERP collection systems usually check whether a request succeeded and whether the HTML parsed. An error triggers a retry. A captcha flags an access problem. A poisoned page passes both checks: it loads, it parses, and it contains organic listings and SERP features. Collection success and data validity are separate conditions, and only a comparison against the keyword's history tests the second one.
How to tell poisoning from a real ranking drop
| Signal | Poisoned SERP | Real ranking change |
|---|---|---|
| Duration | Usually one SERP, back within three days | Persists across weeks |
| Competitors | Most of the usual page-one sites vanish together | Your site moves while others stay |
| Search Console | Impressions and clicks hold steady | Impressions and clicks fall |
| Page shape | Fewer AI Overviews, more image packs and social sites, short pagination | Normal features for the query |
| Result relevance | Off-topic sites, foreign domains, government pages | Relevant competitors |
"Our page dropped out of the top 10 overnight."
Before treating it as an algorithm update or a penalty, check three things: did the other usual sites on that keyword also disappear, did it come back within three days, and does Search Console show the same drop? If the answers are yes, yes, and no, it was a data problem. Our SEO incident investigation guide walks through the full diagnostic workflow.
Comparing against yesterday alone is fragile, because yesterday's SERP may have been poisoned too. Compare against the keyword's usual page-one sites over several weeks instead.
No single signal is proof. Search Console aggregates real search performance under its own methodology, while a rank tracker records individual observations, so the two will never match exactly. Short-lived changes can also come from genuine volatility, personalization, or shifting search intent. Treat each signal as evidence and look for several pointing the same way.
Preserve the original observation either way: timestamp, keyword, location, device, result URLs, positions, and the captured SERP where your tool allows it. Once results return to normal, that record is the only way to investigate what happened.
A validation workflow for rank tracking data
The answer is not to abandon rank tracking. It is to stop treating every collected ranking as settled fact.
Making uncertainty visible improves reporting. A flagged observation should not appear in a client report with the same confidence as a consistent, corroborated ranking.
What this means for SEO reporting
Daily position snapshots, SERP volatility indexes, and share of voice reports all assume Google returns the same page a real searcher would see. If a meaningful share of collected SERPs is now deliberately misleading, single-day rankings can no longer be read at face value.
- Treat Search Console as ground truth for impressions, clicks, and average position. It reflects real users, not scraped pages.
- Report rolling averages rather than daily positions, so one poisoned SERP cannot swing a monthly report.
- Annotate September 13 to 30, 2026 in dashboards and your change log so nobody chases a phantom drop later.
- Pair rankings with outcomes. Organic sessions and key events in GA4 confirm whether a ranking change affected real traffic. See how to measure SEO success.
- Ask your rank tracking vendor how they detect and filter poisoned SERPs, and whether historical data has been flagged.
Poisoned SERP checklist
- Late September 2026 ranking drops checked against Search Console
- Drops checked for recovery within three days
- Competitor positions reviewed for the same keywords on the same day
- Keywords compared against several weeks of usual page-one sites, not just yesterday
- Dashboards annotated for September 13 to 30, 2026
- Client reports use rolling averages and outcome metrics alongside positions
- Original SERP observations preserved for flagged keywords
- Rank tracking vendor asked about poisoned SERP detection
- Legal, home services, and "near me" keywords given extra scrutiny
Frequently asked questions
- What is a poisoned SERP?
- A complete, normal-looking Google results page served to an automated collector, where most of the keyword's usual sites have been replaced. Nozzle defines it as a SERP missing more than half of a keyword's usual sites, which then return on a later SERP.
- Why is Google poisoning search results?
- Nozzle attributes it to Google's response to AI scraping and grounding, the bulk collection of results to feed AI answers. Google has not confirmed this publicly.
- Did my rankings really drop in late September 2026?
- If the drop lasted one to three days, most competitors vanished at the same time, and Search Console shows no change in impressions or clicks, it was very likely a poisoned SERP rather than a real ranking change.
- Has Google confirmed SERP poisoning?
- No. Nozzle's data documents the anomalies, and its explanation is an inference from timing and behavior. Independent replication would strengthen the finding.
- Does poisoning affect what real users see?
- The study measures pages served to automated collection. Search Console data, which reflects real searches, is the best way to confirm whether actual users were affected.
- Which keywords are most affected?
- "Near me" queries, long-tail keywords, and industries such as legal, home services, and weddings. Branded searches and the pets and automotive industries were least affected.
What the study does not yet establish
Nozzle's dataset is large and its method is specific, which makes the findings worth acting on. Four limits still matter:
- The 8.0% depends on the definition. It counts SERPs missing more than half of a keyword's usual sites that later return. It is not a measure of every Google search.
- One collection source. The data comes from Nozzle's own collection routes and customer datasets. Independent replication by unrelated rank trackers would strengthen it.
- Intent is inferred. The behavior is consistent with a response to AI scraping, but the data cannot confirm Google's intent or exactly how result sets were altered.
- Not every brief drop is poisoning. Real volatility, location differences, and personalization can also produce short-lived changes.
These caveats define what the evidence supports. They do not remove the concern. If other trackers confirm the pattern, the impact reaches beyond rank tracking to every piece of third-party search research built on scraped SERPs.
Key takeaway
The evidence points to Google no longer just blocking automated SERP collection, but answering some of it with believable but wrong results. For SEO teams, that means a single day of rank tracking data is now a claim to verify, not a fact to report. Check it against history, against Search Console, and against outcomes before acting on it.
Source: Nozzle Research, "Poisoned Google SERPs", October 10, 2026, data through October 9, 2026. Nozzle publishes the underlying daily, site, country, industry, and SERP feature data as downloadable CSVs. Additional analysis via Search Engine Journal. Hat tip to Cyrus Shepard for surfacing the study.