Your rank report says you fell off page one overnight. Before you rewrite a page, check the report.
As of early November 2026, research from one rank-tracking company shows tracked Google results coming back wrong far more often than they did in August. Since 13 September, page one or two was poisoned on 26.9% of tracked results pages, against 0.28% before. That is about 96 times the earlier rate (our estimate: 26.9 ÷ 0.28). On a poisoned page, for keywords where a tracked site is normally in the top 10, the site's own pages missed the top 10 on 63.5% of poisoned pages, against 2.6% before, so about 24 times as often (our estimate: 63.5 ÷ 2.6).
So a sudden fall in a rank report is now much more likely to be a bad reading than it was in August. The worst point so far was the end of October: on 31 October, 64.0% of tracked page one or two results were poisoned.
How much more often are rank readings wrong now?
Poisoned results pages, before and since 13 Sep
Page one on its own shows the same jump. Page one alone was poisoned on 14.3% of tracked results pages since 13 September, against 0.17% before.
If you've seen an 8% figure quoted, that was the early-study share from 13 September to 9 October, not the current one.
None of these percentages is a share of searches made by real people. They describe pages collected through the tracker's own data routes. It's also worth knowing that the company behind the main study sells rank tracking itself.
What does a poisoned results page look like?
Normal page and poisoned page
A results page is the list Google shows for a search. SEO people call it a SERP. In this research, a results page counts as poisoned when fewer than half of the keyword's usual sites appear anywhere on it. The usual sites are those in the top 10 on at least 60% of baseline pages from 20 August to 9 September.
The problem is that a poisoned page doesn't look broken. The study describes complete, normal-looking results pages with the results swapped out, and says "nothing on the page gives it away".
What changes is who is on the page. Only 29.1% of the previous page's organic sites were kept on a poisoned page, against 88.2% before. The clearest tell is what fills the gap: Social sites, Facebook most of all, take the place of the sites that were there before.
Semrush describes two forms: the whole page is clearly off, or it looks normal at first but includes loosely matching pages further down. There is a smaller sign too. In the early study, an AI Overview showed on 34.9% of poisoned pages, against 54.2% at baseline.
Why are rank trackers getting bad results?
As of 27 October, Google hadn't said. Semrush's 27 October statement says Google "hasn't commented publicly". It calls the issue "industry-wide". Semrush says industry discussion points to "a likely driver": efforts to limit large-scale data collection for AI systems. That is a reading, not a statement from Google.
Barry Schwartz reported a three-part approach as described by Nozzle's Derek Perkins: wide IP blocking, more captchas, and bad data injected into suspect results pages. He added that he doesn't think the target is the SEO tools but AI bots and scrapers.
Whatever the motive, the effect on your report is the same. A tool can record a page that never reflected where you really stand.
When did rank reports start going wrong?
From 20 Aug to 1 Nov 2026
Baseline period for "usual sites" (Nozzle)
Requests to Google start failing (Nozzle)
First poisoned pages (Nozzle)
Google's September 2026 spam update, starting after the poisoning began (Search Status Dashboard via Ryan Doser)
Nozzle publishes its study
Unconfirmed ranking volatility reports
"keeps getting hotter", unconfirmed (Search Engine Roundtable)
Peak day, 64.0% of tracked page 1 or 2 results poisoned (Nozzle)
Unexplained click spike reported in Search Console for some sites (Search Engine Roundtable)
Nozzle updates its data
The timing matters. Google's September 2026 spam update ran from 24 September to 8 October, per Google's Search Status Dashboard as relayed by Ryan Doser, and the poisoning began before it. Not every drop in those weeks was the update.
October was noisy as well. Unconfirmed ranking volatility was reported on 16, 20, 21, 24 and 27 October, Barry Schwartz's 29 October update said things "keep getting hotter", and Google had confirmed nothing as of 27 October. The 31 October peak came two days later.
Does a bad reading fix itself?
Poisoned pages with the usual sites back
Usually, but more slowly than before. 46.9% of poisoned pages had the usual sites back within 3 days, and 90.4% within two weeks. Through 21 October the 3-day share was 67.9%, so spells are lasting longer.
In the early data most spells were brief: 80% lasted one collected page, 15% two, and 5% three or more.
Page two needs extra care. Page two is poisoned about 1.5 times as often as page one, mostly under a page one that looks fine. If you sit between positions 11 and 20, treat a sudden loss there with extra suspicion.
Are Australian, local and mobile results affected?
Share of results pages poisoned, 13 Sep to 9 Oct (third-party analysis of Nozzle's public data)
Australia is not exempt. A third-party analysis of Nozzle's public data from 13 September to 9 October found 11.5% of Australian results pages poisoned, among the highest of the markets with daily data, while the UK was lowest at 3.0%. The same analysis puts India at 10.9% and Canada at 10.5%. On 29 September about 47.5% of Australian pages were poisoned, against 2.8% in the UK.
| Tracked as | Poisoned | Source |
|---|---|---|
| Device (page-1 cut) | ||
| Desktop | 14.9% | Nozzle |
| iPhone | 14.1% | Nozzle |
| Android | 13.0% | Nozzle |
| Same phrase tracked two ways (early-study figure, 21,508 pairs) | ||
| National | 7.5% | Search Engine Journal |
| Local | 7.4% | Search Engine Journal |
Tracking locally doesn't protect you either. The same phrase tracked nationally and locally came out at 7.5% and 7.4% across 21,508 pairs, an early-study figure. The device barely matters: 14.9% on desktop, 14.1% on iPhone and 13.0% on Android.
My rankings dropped: what should I check first?
Before you change a page
Work through these in order. Most take a few minutes.
- Wait for a second reading. Most early spells lasted a single collected page. One vendor says it may keep showing the previous rank for up to two daily updates after an unusually large drop, until the lower rank is confirmed. That is that vendor's rule, not every tool's.
- Compare clicks and impressions in Search Console for the same search and the same page. Semrush's advice for unexpected ranking changes, "especially for keywords unrelated to your pages", is to compare against your own data such as Google Search Console. Remember that average position there is based on the highest-ranking result from your site.
- Use the long view. Set the Performance report to "Last 16 months" to rule out yearly seasonality, and compare the drop period with a similar period.
- Check the Pages table. It shows whether the drop is site-wide, a group of pages or one page.
- Count the keywords. One phrase falling is a different story from dozens. Watch for drops on searches that have nothing to do with your pages.
- Check the Data Anomalies page. Google lists known Search Console data issues there.
- Ask your tool, and re-pull old exports. Some vendors say they've reprocessed affected keywords or are cleaning up past results, but none cites independent testing of its checks. Older exports may no longer match the tool's current numbers if those dates were revised, so compare before you send a report on.
- Don't settle it with one manual search. Results vary by device, location, time and search history.
If Search Console is new to you, our guide to Search Console's AI report and the opt-out switch walks through another part of the same tool.
Which numbers can I trust now?
Rank report shows a drop
- A
Clicks and impressions steady
Small position moves can happen at any time. Avoid radical changes; re-check after a repeat reading.
- B
Impressions hold, clicks fall
Look at page titles and snippets.
- C
Clicks and impressions both fall and stay down
Work through common causes: technical issue, seasonality, algorithm update.
- D
Drop in top results across a wide range of terms
Self-assess the whole site; changes can take days to months to show, so wait a few weeks before re-checking.
Search Console is Google's own data and the better cross-check, but not a perfect one.
Google's own guide is blunt about position: "Impressions and clicks are ultimately the measure of success for your site." It also says small moves can happen at any time "(including moving back up in position, without you needing to do anything)". It recommends "avoiding making radical changes if your page is already performing well".
Read the pattern, not the single number:
- Both clicks and impressions fell. Work through the common causes, such as a technical issue, seasonality or an algorithm update.
- Impressions held but clicks fell. Look at your page titles and snippets.
- Your top results dropped across a wide range of searches. Google says to self-assess the whole site. Changes can take days to months to show, so wait a few weeks before re-checking.
Search Console is Google's own data and the better cross-check, but it isn't perfect right now either. An unexplained spike in reported clicks has shown up for some sites since around 31 October, in some cases in the API but not the web interface, and at least one site owner says it hasn't shown up as traffic on their site. So set your Search Console numbers beside your own site analytics and your enquiries.
At SEO Market, our view is simple. A position is a smoke alarm, not a verdict. Clicks, impressions and real enquiries tell you whether something changed. Don't rewrite a page over a one-day drop. Do dig in when clicks and impressions fall and stay down.
The figures in one place
- 26.9%(before 0.28%)Page 1 or 2 poisoned since 13 SepNozzle
- 64.0%Peak day 31 OctNozzle
- 63.5%(before 2.6%)Tracked site's pages missing the top 10 on poisoned pages, for keywords where the site is a usual top-10 resultNozzle
- 29.1%(before 88.2%)Previous page's organic sites keptNozzle
- 46.9% / 90.4%Usual sites back within 3 days / 14 daysNozzle
- 11.5%(UK 3.0%)Australia, 13 Sep to 9 OctRyan Doser
These are the numbers worth keeping beside your next rank report. They come from tracked pages, not from real people's searches, and the live study may still revise them.
Not sure whether a drop is real?
Pull up the search term, the page and the dates the report flagged. Check them against Search Console and your enquiries before anyone touches the page. If you'd like a second pair of eyes before you change content, get in touch with SEO Market through our contact page.
Sources
- https://nozzle.io/research/poisoned-google-serps/
- https://pikaseo.com/articles/google-poisoned-search-results-rank-trackers-2026
- https://ryandoser.com/research/poisoned-google-serps/
- https://www.searchenginejournal.com/analysis-google-is-poisoning-search-results-for-rank-trackers/592873/
- https://www.searchenginejournal.com/rank-trackers-suspect-google-results/594049/
- https://www.semrush.com/blog/poisoned-serps/
- https://www.seroundtable.com/third-party-google-rank-tracking-broken-42272.html
- https://www.seroundtable.com/google-search-ranking-volatility-42379.html
- https://www.seroundtable.com/google-performance-reports-spike-clicks-42448.html
- https://developers.google.com/search/docs/monitor-debug/debugging-search-traffic-drops