Google Serves ‘Poisoned’ Search Results to Rank Trackers, Nozzle Reports

What Nozzle found

Nozzle examined a sample of the Google SERPs it collected between August 20 and October 9, 2026. The sample spans 49 countries, desktop and mobile devices, and different location targets. The company reports more than 660,000 keywords with at least one poisoned capture. Its headline rate is 8.0% of eligible SERPs since September 13, compared with a 0.17% baseline before the disruption.

Its timeline has two stages. Many automated requests began failing on September 13. The first SERPs meeting its poisoning test appeared on September 20. Daily rates peaked near 27% on September 28 and 29, then ranged from roughly 5% to 11% on most days from October 1 through October 9. The last days can change as later captures establish whether familiar sites returned.

All figures: Nozzle, data through October 9, 2026. The percentages have different denominators.

How the study defined a “poisoned” page

Nozzle first built a usual result set for each keyword: sites appearing in the top 10 on at least 60% of five or more baseline captures from August 20 to September 9. A later SERP qualified as poisoned when fewer than half of those usual sites appeared anywhere in its pages and most came back on a subsequent capture. The full method and downloadable aggregates are public.

That return requirement matters. A real ranking change can remove familiar sites too; the study tries to isolate abrupt, temporary replacement of the entire result set. It also means the measure is retrospective. You cannot label a brand-new suspect page with certainty until more observations arrive.

The captured page still has normal Google page structure rather than an error or CAPTCHA. A tracker that accepts a successful response and parses positions can store the swapped listings as if they were ordinary rankings. That is why an apparently precise position chart may be less trustworthy than a visibly failed fetch.

What an altered SERP looks like

Nozzle shows the U.S. desktop query “white label local seo” on October 4 and October 7. Its normal page listed white-label SEO vendors. Three days later, all nine sites from page one were absent from every page of the altered capture. Results included a WordPress development product, a restaurant, and government pages matching pieces of the words “local” and “SEO” rather than the searcher's likely intent. Pagination shortened from 10 pages to three.

Across the study, a tracked project's own site was outside the top 10 on 58.6% of poisoned SERPs, compared with 2.7% during the baseline. Reddit appeared in the top 10 more often on poisoned pages for the same keywords, rising from 20.8% to 28.8%; Facebook rose from 28.3% to 33.8%. These shifts describe the pages Nozzle collected. They are not evidence that human searchers suddenly saw those sites rise.

The disruption also reaches beyond blue links. Nozzle found less continuity in People Also Ask and AI Overview sources on poisoned captures. That raises a measurement question for tools that track AI citations from scraped SERPs, even when a traditional rank report is not their main product.

What the evidence does and does not show

The evidence is unusually detailed for a vendor report: Nozzle publishes its rule, daily counts, country and device cuts, examples, and aggregate downloads. Its data is still drawn from Nozzle's collection routes and customer datasets. The 8.0% rate is not an estimate of all Google searches, every scraper, or every rank-tracking provider. Nozzle also sells rank-tracking software, so independent replication would strengthen the conclusion.

Nozzle attributes the pattern to Google serving altered results to automated collectors. The observed pages establish a serious integrity problem for those captures; they do not, on their own, prove Google's intent or reveal its detection logic. We have not found a public Google confirmation of this specific behavior. The safest phrasing for a client report is that Nozzle observed and classified altered SERPs in its rank-tracking data.

There is also a separate, previously reported pressure on SERP collection: Google's /goto result links made some links harder for tools to resolve. That changes collection cost and parsing. The October study describes a different failure mode: the result page itself can look valid while carrying unexpected listings. Neither event proves your website lost visibility.

How to check a reported ranking drop

A one-day collapse is a signal to investigate, not a reason to rewrite a page. Use the same query, page, country, device, and time window wherever possible:

1. Save the tracker evidence. Export the affected keywords, timestamps, target locations, devices, and any stored SERP HTML or screenshots. Mark the first abnormal capture and the next normal one.
2. Compare first-party performance. In Search Console's Performance report, filter the relevant query and page, then inspect clicks, impressions, and average position across several days. Google's traffic-drop guide recommends looking for sustained changes and affected page groups, with clicks and impressions carrying more weight than an isolated absolute position.
3. Inspect the result set. Compare the tracker's captured page with another capture for the same query, country, device, and location. A browser check can help, but personalization, location, and timing mean one manual search is not a perfect control.
4. Wait for a repeat capture. Nozzle says 80% of poisoning spells lasted one SERP and 77.9% of eligible cases had their usual sites back within three days. A quick reversal supports a collection issue; it is not proof by itself.
5. Escalate when first-party signals agree. If Search Console clicks and impressions also fall and the decline persists, investigate the affected pages, demand, indexing, and any recent site changes. A genuine decline can occur during the same period as unreliable tracker data.

What to ask your rank tracker

Ask your vendor how it detects a complete but implausible SERP, whether it compares each keyword with a multiweek baseline, and whether it automatically rechecks outliers before publishing them. Ask whether historical dates were flagged or reprocessed, and whether you can download the raw capture behind a position change. A clear answer is more useful than a blanket claim that the tool is unaffected.

For client dashboards, annotate the late-September and early-October 2026 collection period if your provider had anomalies. Show Search Console clicks and impressions beside tracked positions. Keep rank tracking for competitive context and query-level monitoring, but make the measurement confidence visible when the source page itself is in doubt.

The next useful evidence would be independent studies from other providers, a Google explanation, and transparent vendor incident notes. Until then, Nozzle's study is a strong reason to verify sudden rank losses carefully, not a reason to dismiss every decline as a scraper artifact.

Frequently asked questions

Sources

• Nozzle: Poisoned Google SERPs, study and methodology
• Google: Debug Google Search traffic drops
• Google: Search Console Performance report