Your rank tracker shows a drop. Is it real or a poisoned SERP?

KEY TAKEAWAYS

1. A one-day drop in a rank tracker is a lead, not a finding. Most poisoned spells last one check.
2. At an 8% poisoning rate, a single top-10 miss has roughly a one-in-three to two-in-three chance of being bad data.
3. Search Console is the referee for your own site, with known limits on lag, time zone and how position is averaged.
4. Ask your vendor how it detects swapped pages. Restored depth is not the same as detection.

Since late September 2026, a share of the Google results pages collected by rank trackers have been wrong. They are not error pages. They look complete, but most of the sites that normally rank for the keyword are gone. A tracker stores that page as data, and your report shows a drop that never happened in Search. Some drops in the same weeks were real. This article sets out how to tell them apart.

What is a poisoned SERP?

Nozzle, an enterprise rank tracker, published the main study on 10 October. It defines a poisoned page by comparison with each keyword's history. A keyword's usual sites are those in its top 10 on at least 60% of its pages between 20 August and 9 September. A later page is poisoned when fewer than half of those sites appear on it and they come back soon after.

The page carries no captcha, error code or marker. That is the difference from blocking. A blocked request fails and the tool retries. A swapped page succeeds and gets saved.

The sequence matters. Google removed the num=100 parameter in September 2025, so tools had to page through results ten at a time. In late August 2026 Google confirmed it was routing result links through google.com/goto redirects. Nozzle's founder put the cost of resolving one five-page ranking at 500 to 1,000 requests. From 13 September, requests began failing in bulk, and he told Search Engine Roundtable retrievable data fell about 80%. Swapped pages first appeared on 20 September.

One naming note. "Poisoned SERP" has an older meaning: malicious pages shown to real searchers. This is a different problem. No source shows that ordinary users see the swapped pages. The evidence covers automated collection only.

How big is the problem?

The published figures do not form a range. Each counts something different.

Nozzle's own data shows why vendors disagree. On the same days, one of its collection routes returned 1% to 7% poisoned pages and another 9% to 30%. Your exposure depends on how your vendor collects, not on a single industry number.

It also depends on what you track. Nozzle found Australia at 11.5%, the US at 8.5% and the UK at 3.0%. US legal keywords ran 15.9% across the period. "Near me" searches ran 11.4% against 7.8% for the rest. The most navigational brand phrases were lowest, at 5.3%. Device made little difference.

Was it the spam update instead?

Google ran its September 2026 spam update from 24 September to 8 October. SE Ranking linked its users' volatility to that kind of update plus Google's collection changes. Nozzle's study does not discuss the update at all. Both views are partly right.

The timeline rules out the update as the main cause. Ryan Doser recomputed Nozzle's published files and found the rise began four days before the update. Elsop counted 43,644 poisoned pages between 20 and 23 September. But Elsop also found 93.6% of all poisoned pages fell inside the update window, which held 68.0% of judged pages. The two events overlapped almost completely.

There is a second point that settles part of the argument. Nozzle's definition requires the usual sites to come back. A loss that persists is excluded by design. So the 8.0% is not "drops that were fake out of all drops." Real update losses sit outside it. If you are diagnosing the update itself, start from our September 2026 spam update diagnosis and use Search Console data, not tracker data, for that window.

Did the tools fix it?

Several vendors say yes. Ranktracker says it restored full top 100 tracking on 1 October. SE Ranking says most missing positions were resolved. TrueRanker says its crawls were collecting organic positions correctly again by 16 October. SISTRIX noted on 26 September that Google was sometimes returning incomplete or distorted pages, and it added a quality check that updates fewer keywords a day.

Read these claims carefully. Most describe depth and completeness: getting all the results back. A tracker can return a full page of ten results and still store a swapped page. TrueRanker makes exactly this point. Nozzle's daily rate of 5% to 11% after 1 October came after most of these notices.

Tools also behaved differently. Glenn Gabe compared them on a site that lost traffic after the update. Ahrefs caught the drop, Semrush caught it later, and SISTRIX had not shown it by 7 October. SE Ranking says it cannot correct its data from 11 to 16 September.

Do not expect Google to change course. Its spam policies class "scraping results for rank-checking purposes" without permission as machine-generated traffic that violates them. Google has not commented on the swapped pages.

What are the odds a single drop is bad data?

Nozzle's figures let you estimate this. On poisoned pages the tracked site was out of the top 10 58.6% of the time. On normal pages it was out 2.7% of the time in the baseline, and 10.2% on the same keywords' other pages since 13 September. That second figure includes real update losses, so it sets an upper bound for real misses.

Our calculation applies Bayes' rule with those rates. The chance that a top-10 miss comes from a poisoned page equals the poisoning rate times 58.6%, divided by that amount plus the normal miss rate times the remaining pages.

The lower figure in each range uses the 10.2% normal miss rate and the higher uses 2.7%. At typical rates, a single miss is close to a coin toss. That is too uncertain to justify a content rewrite or a client escalation.

Time narrows it fast. Nozzle found 80% of poisoned spells last one check, 15% last two and 5% last three or more. A miss that holds for three checks in a row has outlasted 95% of spells. A real loss tends to persist.

How should you check a drop?

Use this order. It costs a few minutes per keyword and avoids most false alarms.

1. Record the reading. Note keyword, URL, position, date, country and device. Open the tool's saved copy of the results page if it keeps one. SE Ranking advises this: a short or unfamiliar page with your site missing points to collection.
2. Compare with weeks, not yesterday. Check whether at least half of the keyword's usual top 10 sites are on the saved page. Nozzle warns that yesterday's page may also be swapped. Reddit, Facebook and LinkedIn gained share on poisoned pages, so a sudden wall of social results is a hint, not proof.
3. Wait for three checks. Nozzle found 77.9% of poisoned pages back to normal within three days, with a median of two.
4. Ask Search Console. Filter the Performance report to the same query and page, country and device. Look at impressions, clicks and position for the days in question.
5. Read Search Console with its limits. Position is the topmost spot your site held, averaged across searches, so it will not match one tracker reading. Data usually lands in two to three days, days are labelled in Pacific Time, and the newest data is preliminary. The report also ran over 64 hours late on 11 October. Rare queries may be anonymized and leave no row at all.
6. Escalate only on agreement. If the drop holds past three checks and Search Console shows the same query and page losing position or impressions, treat it as real. Then check indexing, manual actions, competitors and the update timeline.
7. Ask your vendor three questions. Which collection route serves your project? Does it detect swapped pages or only missing ones? Does it keep or discard suspect pages?

One caution on step 4. Search Console impressions are not entirely free of bot effects. When num=100 went away, desktop impressions fell, and Zeo and others read that as scraper impressions dropping out. Clicks and position are the steadier signals.

What does this look like in practice?

A hypothetical example. A US law firm tracks 400 keywords daily. On 28 September its tracker shows 36 keywords falling out of the top 10 overnight.

By 1 October, after three checks, 27 are back in their usual positions. Those were transient and need no action. Of the 9 still missing, Search Console shows 6 with impressions at an average position under 10 on the same days. The tracker is wrong on those. The other 3 show position slipping past 14 and impressions falling. Those are the real losses, and they fall inside the spam update window, so they get a full diagnosis.

The report says 3 confirmed losses, not 36. Keywords with no Search Console rows stay on a watch list, marked unverified.

What should you report until this settles?

Report your own rankings from Search Console. It is generated inside Google and does not depend on scraping. Use the tracker for what Search Console cannot show: competitor positions and the layout of the results page.

Treat tracker-based SERP feature data with extra care. AI Overviews appeared on 34.9% of poisoned pages against 54.2% normally. On 27 August, DataForSEO reported at least one unresolved goto link on 52.26% of the pages where it found an AI Overview. Any AI Overview presence series collected in these weeks may undercount. That includes branded-query tracking like the work behind our AI Overviews on branded searches audit. For AI surfaces, Google's own reporting is the cleaner baseline, with the limits set out in our note on the Search Console generative AI report.

Mark every tracker change in a report as provisional until it passes three checks and Search Console agrees. If you want to see how your brand appears in search and AI answers without relying on scraped data alone, join the waitlist.

Questions

What is a poisoned SERP?

It is a complete, normal-looking Google results page, served to an automated request, in which most of a keyword's usual sites are missing. Nozzle counts a page as poisoned when fewer than half of the usual top 10 sites appear and they come back within days. Nothing on the page signals the swap.

How common are poisoned SERPs?

It depends on the vendor and route. Nozzle measured 8.0% of judged pages from 13 September to 9 October 2026, with a 27.0% peak on 29 September. DemandSphere reported about 11% of its daily pages. Nozzle saw 1% to 7% on one collection route and 9% to 30% on another on the same days.

Was the drop caused by the September 2026 spam update instead?

Possibly both. Swapped pages began on 20 September, four days before the update started on 24 September, so the update does not explain them. But 93.6% of Nozzle's poisoned pages fell inside the update window. Use Search Console, not tracker data, to judge the update for that period.

Has Google said anything about it?

Not about swapped pages. Google's only statement on recent changes, given for the August goto redirects, was that it has a long history of technical measures against abuse. Its spam policies also list scraping results for rank checking without permission as a violation, so there is no reason to expect a fix.

Sources

1. Nozzle, Poisoned Google SERPs nozzle.io
2. Ryan Doser, Poisoned Google SERPs analysis, data through October 9, 2026 ryandoser.com
3. Elsop, Rank trackers get poisoned Google results. What 8.0% means elsop.com
4. DemandSphere, Spurious data in Google SERPs demandsphere.com
5. TrueRanker, Google is serving rank trackers pages that look real and aren't trueranker.com
6. Search Engine Roundtable, Third-party Google rank tracking not fully working now seroundtable.com
7. Search Engine Roundtable, Google may be more successful in blocking scrapers and tracking tools seroundtable.com
8. Search Engine Journal, Analysis. Google is poisoning search results for rank trackers searchenginejournal.com
9. PPC Land, Google returns wrong results on 11% of DemandSphere's daily rank checks ppc.land
10. PPC Land, Google forces rank trackers into 500 to 1,000 requests per query ppc.land
11. On-Page.ai, google.com/goto is live. What the black-box SERP breaks for rank trackers blog.on-page.ai
12. Google Search Central, Spam policies for Google web search developers.google.com
13. Search Console Help, What are impressions, position, and clicks support.google.com
14. Search Console Help, Performance report (Search results). Troubleshooting data discrepancies support.google.com
15. Search Console Help, Performance report (Search results) support.google.com
16. Search Engine Roundtable, Google Search Console performance report extremely delayed seroundtable.com
17. Search Engine Land, Google September 2026 spam update done rolling out searchengineland.com
18. SISTRIX, Status page status.sistrix.com
19. SE Ranking, Issue resolved. Inaccurate positions in Rank Tracker help.seranking.com
20. SE Ranking, Google rankings delay. What happened? help.seranking.com
21. National Law Review, Ranktracker restores Google Top 100 rankings as SEO platforms face widespread disruptions natlawreview.com
22. Zeo, The impact of num=100 parameter removal on SEO reporting zeo.org

NEXT STEP

See where your brand stands.

Your Digital Authority Score, who AI recommends for your target questions, and the placements that close each gap. Your first report is free.

Get your free report →