Nozzle Research
Poisoned Google SERPs
Since September 2026, Google has been poisoning the SERPs it serves rank trackers: complete, normal-looking pages whose results have been swapped out. Nothing on the page gives it away. Each keyword’s own history does.
- SERPs analyzed
- 29.9M
- poisoned on page 1 or 2
- 22.6%
- poisoned at the peak, September 30
- 43.5%
- as often poisoned on page 2 as on page 1
- 1.5×
Key findings
- 22.6% poisoned on average, 43.5% at the peakSERPs with page 1 or page 2 poisoned since September 13, of those judged on both pages, against 0.28% before. It eased to 4 to 6% in mid-October, then climbed back to 35.7% on October 21.
- Page 2 is hit harder, in two waysPage 2 is poisoned one and a half times as often as page 1, mostly under a page 1 that looks fine, and for two weeks it answered one word of the query instead.
- Blocked first, then poisonedRequests to Google started failing on September 13; from September 20, the ones that got through came back poisoned.
- The page looks normal. The results don’t.The organic results keep 22.3% of the previous SERP’s sites, against 88.1% before, and social sites and forums take their place.
- Tracked sites drop out, then come backA project’s own site misses the top 10 on 60.8% of poisoned SERPs, against 2.7% before, and 67.9% have their usual sites back within three days.
The search “white label local seo” loads page 1, and three days later every site on it has been replaced, with nothing on the page to say so; a rank report reads it as a crash. The film then follows the study: the sample’s SERPs since September 13, when requests to Google started failing, poisoned SERPs from September 20, why a poisoned SERP is worse than a blocked request, the social sites and forums that take the usual sites’ place, tracked sites falling out of the top 10, and how quickly it passes.
Update, October 21: page 2
The study now counts a SERP as poisoned when page 1 or page 2 is: 22.6% since September 13, and 43.5% on September 30, on the SERPs judged on both pages. Page 2 is hit harder: poisoned one and a half times as often as page 1, three times in five under a page 1 that looks fine, and for two weeks in September it answered one word of the query instead. By the numbers has the chart.
The poisoning eased on October 14 and 15, to 4 to 6% of SERPs, then came back: 25.9% on October 19 and 35.7% on October 21, the worst day since October 8.
What it looks like
Each pair is the same keyword, device and target location, one to three days apart. Pick one, then drag the divider across the page.
Drag the divider, or click anywhere on the page to move it there. Scroll inside the frame to move down both SERPs together.
By the numbers
A page is poisoned when fewer than half of its keyword’s usual sites for that page, the ones on it in at least 60% of its SERPs from August 20 to September 9, appear anywhere on the SERP. A SERP counts as poisoned when page 1 or page 2 is.
On September 13, requests to Google started failing in large numbers. The ones that got through were normal until September 20, when poisoned SERPs first appear. That sequence is why we call it poisoning: Google is not refusing these requests, it answers them with a complete page whose results are wrong. A refused request tells a scraper to try again; a poisoned one gets stored as real. It is Google’s response to AI scraping, and rank trackers are caught in it.
Two things go wrong on page 2. One is poisoning: the page answers the query, and the usual sites are gone. The other is Google ignoring part of the query, often all but one word, as “white label local seo” does in the first example above. SerpApi has tracked that fault since May, and it has a study of its own, Google Plays Dumb. The chart follows both pages of the same desktop SERPs: the thick line is page 1 or 2, and the different-query pages are dashed.
Page 1 or 2 poisoned
0.28% to 22.6%
SERPs poisoned on either page: baseline, then since Sep 13.
Page 1 poisoned
0.04% to 12.2%
SERPs with fewer than half of their usual sites: baseline, then since Sep 13.
Page 2 poisoned
0.25% to 17.9%
The same SERPs, with fewer than half of their usual page-2 sites: baseline, then since Sep 13.
Under a page 1 that looks fine
58.4%
Of the SERPs poisoned on page 2 since Sep 13, the share whose page 1 still had its usual sites.
Poisoned SERPs, per day, by page
Thick line: SERPs poisoned on page 1 or page 2. Solid lines: each page poisoned, answering the query sent with the usual sites gone. Dashed lines: pages that ignore part of the query. Shaded: the baseline, Aug 20 to Sep 9. Vertical dashes: Sep 13, when requests to Google started failing, and Sep 20, when poisoned SERPs first appear on page 1. Hollow points: thin days, with fewer than 25,369 SERPs (a fifth of a usual day), where a few SERPs move the line a lot; a break is a day with fewer than 100.
Daily table
| Day | SERPs | Page 1 or 2 poisoned | Page 1 poisoned | Page 2 poisoned | Both | Page 1 words ignored | Page 2 words ignored |
|---|---|---|---|---|---|---|---|
| Aug 20 | 210,577 | 0.24% | 0.05% | 0.21% | 0.02% | 0.01% | 0.04% |
| Aug 21 | 12,667 | 0.26% | 0.06% | 0.21% | 0.02% | 0.0% | 0.03% |
| Aug 22 | 5,258 | 0.30% | 0.13% | 0.21% | 0.04% | 0.04% | 0.02% |
| Aug 23 | 32,223 | 0.17% | 0.03% | 0.16% | 0.01% | 0.01% | 0.01% |
| Aug 24 | 345,247 | 0.40% | 0.06% | 0.35% | 0.01% | 0.02% | 0.06% |
| Aug 25 | 416,774 | 0.37% | 0.05% | 0.33% | 0.01% | 0.02% | 0.07% |
| Aug 26 | 361,012 | 0.29% | 0.03% | 0.26% | 0.01% | 0.01% | 0.03% |
| Aug 27 | 124,486 | 0.22% | 0.03% | 0.19% | 0.00% | 0.01% | 0.04% |
| Aug 28 | 433,677 | 0.25% | 0.03% | 0.22% | 0.01% | 0.01% | 0.04% |
| Aug 29 | 480,264 | 0.24% | 0.03% | 0.21% | 0.00% | 0.02% | 0.04% |
| Aug 30 | 408,406 | 0.28% | 0.04% | 0.25% | 0.01% | 0.02% | 0.05% |
| Aug 31 | 513,345 | 0.24% | 0.03% | 0.21% | 0.01% | 0.01% | 0.04% |
| Sep 1 | 409,213 | 0.28% | 0.04% | 0.25% | 0.01% | 0.02% | 0.03% |
| Sep 2 | 511,537 | 0.28% | 0.06% | 0.25% | 0.03% | 0.02% | 0.04% |
| Sep 3 | 376,103 | 0.27% | 0.03% | 0.24% | 0.01% | 0.02% | 0.06% |
| Sep 4 | 1,484 | 0.20% | 0.07% | 0.13% | 0.0% | 0.0% | 0.13% |
| Sep 5 | 1,412 | 0.42% | 0.21% | 0.28% | 0.07% | 0.0% | 0.0% |
| Sep 6 | 1,388 | 0.50% | 0.22% | 0.43% | 0.14% | 0.07% | 0.14% |
| Sep 7 | 1,414 | 0.57% | 0.35% | 0.35% | 0.14% | 0.0% | 0.0% |
| Sep 8 | 1,434 | 0.28% | 0.07% | 0.21% | 0.0% | 0.0% | 0.0% |
| Sep 9 | 8,228 | 0.33% | 0.15% | 0.26% | 0.07% | 0.01% | 0.17% |
| Sep 10 | 378,655 | 0.42% | 0.12% | 0.36% | 0.06% | 0.04% | 0.09% |
| Sep 11 | 0 | – | – | – | – | – | – |
| Sep 12 | 9 | – | – | – | 0.0% | – | – |
| Sep 13 | 391 | 1.0% | 0.26% | 1.0% | 0.26% | 0.0% | 0.0% |
| Sep 14 | 7,419 | 0.30% | 0.03% | 0.27% | 0.0% | 0.09% | 0.08% |
| Sep 15 | 8,532 | 0.38% | 0.07% | 0.36% | 0.06% | 0.11% | 0.15% |
| Sep 16 | 73,543 | 1.2% | 0.07% | 1.1% | 0.02% | 0.03% | 5.0% |
| Sep 17 | 44,819 | 4.3% | 0.16% | 4.1% | 0.03% | 0.04% | 22.7% |
| Sep 18 | 95,973 | 4.2% | 0.19% | 4.0% | 0.03% | 0.03% | 23.4% |
| Sep 19 | 139,204 | 3.7% | 0.20% | 3.6% | 0.04% | 0.04% | 20.4% |
| Sep 20 | 137,616 | 4.1% | 0.19% | 3.9% | 0.05% | 0.09% | 22.1% |
| Sep 21 | 64,065 | 4.4% | 0.23% | 4.2% | 0.07% | 0.07% | 23.7% |
| Sep 22 | 206,399 | 6.0% | 0.91% | 5.5% | 0.38% | 0.19% | 29.0% |
| Sep 23 | 144,389 | 9.2% | 2.8% | 7.7% | 1.3% | 0.44% | 28.7% |
| Sep 24 | 114,653 | 11.5% | 4.5% | 9.1% | 2.1% | 0.58% | 30.3% |
| Sep 25 | 147,442 | 18.3% | 9.3% | 13.5% | 4.5% | 1.1% | 28.4% |
| Sep 26 | 278,467 | 26.9% | 15.8% | 18.7% | 7.6% | 1.7% | 30.9% |
| Sep 27 | 188,773 | 34.9% | 21.9% | 23.4% | 10.5% | 2.6% | 35.0% |
| Sep 28 | 56,247 | 35.7% | 23.4% | 22.0% | 9.7% | 1.5% | 39.3% |
| Sep 29 | 142,283 | 38.2% | 23.2% | 28.1% | 13.2% | 2.6% | 27.2% |
| Sep 30 | 134,757 | 43.5% | 24.4% | 38.4% | 19.3% | 2.7% | 3.9% |
| Oct 1 | 114,898 | 31.6% | 17.5% | 27.7% | 13.6% | 2.1% | 3.2% |
| Oct 2 | 135,961 | 20.3% | 10.7% | 17.0% | 7.3% | 1.4% | 1.8% |
| Oct 3 | 107,567 | 38.5% | 22.2% | 31.9% | 15.7% | 3.8% | 4.3% |
| Oct 4 | 122,434 | 38.3% | 21.5% | 31.6% | 14.8% | 3.4% | 4.0% |
| Oct 5 | 103,993 | 40.4% | 22.9% | 33.7% | 16.2% | 4.1% | 4.7% |
| Oct 6 | 80,890 | 35.8% | 18.1% | 29.2% | 11.6% | 3.5% | 3.6% |
| Oct 7 | 157,995 | 39.5% | 21.3% | 32.9% | 14.7% | 4.1% | 4.6% |
| Oct 8 | 156,825 | 35.9% | 18.9% | 29.2% | 12.3% | 3.9% | 3.9% |
| Oct 9 | 123,933 | 34.6% | 19.5% | 27.2% | 12.1% | 4.9% | 4.2% |
| Oct 10 | 138,173 | 33.3% | 18.2% | 26.0% | 10.9% | 4.5% | 3.7% |
| Oct 11 | 122,305 | 34.1% | 19.2% | 26.8% | 11.8% | 5.0% | 4.2% |
| Oct 12 | 152,662 | 27.1% | 14.9% | 21.1% | 8.9% | 3.1% | 2.8% |
| Oct 13 | 117,776 | 31.7% | 17.6% | 24.8% | 10.8% | 4.6% | 4.0% |
| Oct 14 | 168,121 | 5.7% | 2.4% | 4.7% | 1.4% | 0.66% | 0.69% |
| Oct 15 | 147,915 | 3.8% | 1.3% | 3.2% | 0.74% | 0.48% | 0.55% |
| Oct 16 | 158,198 | 8.3% | 3.8% | 6.6% | 2.1% | 0.80% | 0.74% |
| Oct 17 | 126,843 | 14.6% | 7.6% | 11.5% | 4.6% | 2.1% | 1.8% |
| Oct 18 | 174,557 | 14.5% | 7.6% | 11.4% | 4.5% | 2.0% | 1.6% |
| Oct 19 | 102,981 | 25.9% | 14.7% | 20.3% | 9.1% | 3.4% | 2.9% |
| Oct 20 | 111,659 | 20.1% | 10.7% | 15.8% | 6.3% | 2.7% | 2.2% |
| Oct 21 | 65,390 | 35.7% | 20.0% | 28.0% | 12.3% | 5.1% | 4.3% |
Download the data (63 rows, one per day; gzipped CSV). It holds every number in the chart and the tiles, with no account or keyword in it.
What each column means
- date
- Day the SERP was scheduled, UTC
- judged
- Desktop SERPs that day whose keyword has usual sites and usual page-2 sites, with five or more organic results on each page
- page1_poisoned
- Of those, SERPs with fewer than half of the keyword’s usual sites on any page, the study’s test, where page 1 answers the query sent
- page2_poisoned
- SERPs with fewer than half of the keyword’s usual page-2 sites on any page, the same test on page 2, where page 2 answers the query sent
- both_poisoned
- SERPs poisoned on both pages; page1_poisoned + page2_poisoned − both_poisoned is the page 1 or 2 line
- page1_other_query
- SERPs whose page 1 ignores part of the query: a word of the query that at least 60% of the keyword’s baseline page-1 results carried is on at most 20% of them
- page2_other_query
- The same for page 2
Page 1 alone covers every SERP of the study: 8.7% poisoned since September 13, against 0.17% before. The rest of the study counts page 1. Pick a country or a device.
Poisoned SERPs
0.17% to 8.7%
Baseline, then every SERP since Sep 13. Daily since then: 0–29%.
Sites gone per SERP
0.23 to 1.75
Of the previous SERP’s top 10, baseline then since Sep 13.
Short pagination
0.19% to 32.5%
Desktop SERPs whose page numbers stop at 4 or earlier, on keywords that usually run to page 8: baseline, then the worst day, Sep 28.
Back within three days
67.9%
Of the poisoned SERPs since Sep 13, the share whose keyword had its usual sites back within three days.
Share of SERPs missing most of the keyword’s usual sites, per day
Shaded: the baseline, Aug 20 to Sep 9. Dashed: Sep 13, when requests to Google started failing and the period the tiles compare begins, and Sep 20, when poisoned SERPs first appear. Hollow points: thin days, with fewer than 198,646 judged SERPs in this slice (a fifth of its usual day), where a few SERPs move the line a lot.
Daily table
| Day | All SERPs | Judged SERPs |
|---|---|---|
| Aug 20 | 0.06% | 937,095 |
| Aug 21 | 0.12% | 1,137,704 |
| Aug 22 | 0.12% | 933,999 |
| Aug 23 | 0.11% | 1,097,483 |
| Aug 24 | 0.06% | 993,230 |
| Aug 25 | 0.05% | 1,089,001 |
| Aug 26 | 0.45% | 1,324,350 |
| Aug 27 | 1.2% | 1,146,448 |
| Aug 28 | 0.04% | 1,100,991 |
| Aug 29 | 0.04% | 1,197,012 |
| Aug 30 | 0.04% | 1,043,040 |
| Aug 31 | 0.04% | 1,252,272 |
| Sep 1 | 0.05% | 948,502 |
| Sep 2 | 0.07% | 1,241,572 |
| Sep 3 | 0.04% | 987,715 |
| Sep 4 | 0.13% | 1,200,368 |
| Sep 5 | 0.13% | 978,110 |
| Sep 6 | 0.15% | 1,142,666 |
| Sep 7 | 0.17% | 1,034,224 |
| Sep 8 | 0.20% | 1,125,779 |
| Sep 9 | 0.18% | 1,018,813 |
| Sep 10 | 0.15% | 1,121,881 |
| Sep 11 | 0.13% | 991,655 |
| Sep 12 | 0.18% | 1,098,064 |
| Sep 13 | 0.18% | 888,364 |
| Sep 14 | 0.34% | 718,036 |
| Sep 15 | 0.18% | 294,126 |
| Sep 16 | 0.08% | 187,238 |
| Sep 17 | 0.17% | 109,952 |
| Sep 18 | 0.19% | 234,441 |
| Sep 19 | 0.19% | 367,975 |
| Sep 20 | 2.0% | 597,749 |
| Sep 21 | 2.5% | 303,965 |
| Sep 22 | 1.6% | 550,114 |
| Sep 23 | 4.1% | 433,705 |
| Sep 24 | 6.4% | 367,449 |
| Sep 25 | 11.6% | 406,244 |
| Sep 26 | 19.6% | 645,700 |
| Sep 27 | 26.3% | 448,890 |
| Sep 28 | 28.2% | 146,156 |
| Sep 29 | 28.6% | 344,855 |
| Sep 30 | 13.8% | 1,233,742 |
| Oct 1 | 8.8% | 1,004,109 |
| Oct 2 | 5.7% | 1,196,122 |
| Oct 3 | 11.1% | 957,652 |
| Oct 4 | 9.8% | 1,117,691 |
| Oct 5 | 12.1% | 1,009,781 |
| Oct 6 | 11.6% | 1,107,417 |
| Oct 7 | 14.5% | 1,002,183 |
| Oct 8 | 10.8% | 1,101,965 |
| Oct 9 | 9.4% | 992,346 |
| Oct 10 | 9.1% | 1,090,276 |
| Oct 11 | 9.6% | 934,913 |
| Oct 12 | 8.2% | 1,142,884 |
| Oct 13 | 8.8% | 921,888 |
| Oct 14 | 3.2% | 1,131,636 |
| Oct 15 | 3.1% | 918,518 |
| Oct 16 | 3.6% | 1,120,708 |
| Oct 17 | 6.6% | 901,679 |
| Oct 18 | 7.0% | 1,036,650 |
| Oct 19 | 11.4% | 954,974 |
| Oct 20 | 9.0% | 1,058,851 |
| Oct 21 | 15.6% | 935,350 |
Download the data (1,701 rows, one per day, country and device; gzipped CSV). It holds every number in the chart and the tiles, with no account or keyword in it.
What each column means
- date
- Day the SERP was scheduled, UTC
- country
- Target country, ISO 3166 code; other is every country under 1% of SERPs, combined
- device
- desktop, iphone or android
- judged
- SERPs whose keyword has usual sites: in its top 10 on at least 60% of five or more baseline SERPs
- degraded
- Poisoned SERPs: judged SERPs with fewer than half of the usual sites on any page
- with_previous
- SERPs with a SERP of the same keyword in the 48 hours before
- previous_top10_missing
- Sum over those SERPs of the previous SERP’s top-10 sites gone from every page
- deep_desktop
- Desktop SERPs of keywords whose baseline pagination ran to page 8 or beyond
- short_pagination
- Of those, SERPs whose page-1 pagination stops at page 4 or earlier, or shows none
- degraded_followed_up
- Poisoned SERPs followed by another SERP of the keyword within three days
- degraded_back_3d
- Of those, SERPs whose keyword had at least 80% of its usual sites back within three days
The device barely matters: 8.8% on desktop, 8.8% on iPhone and 7.7% on Android.
What takes their place
Social platforms and forums. How often a site is in the top 10 of a keyword’s poisoned SERPs, against the same keyword’s other SERPs since September 13.
Share of SERPs with the site in the top 10
- The same keywords’ other SERPs
- Poisoned SERPs
More often on poisoned SERPs
- facebook.com21.9% → 29.5%
- instagram.com6.1% → 10.4%
- youtube.com12.2% → 15.4%
- quora.com2.4% → 5.0%
- reddit.com20.7% → 23.0%
- etsy.com0.7% → 2.8%
- linkedin.com3.7% → 5.4%
- mapquest.com2.5% → 3.4%
- tiktok.com0.7% → 1.6%
- justanswer.com1.4% → 2.0%
Less often on poisoned SERPs
- yelp.com11.5% → 10.8%
Since Sep 13, each keyword’s poisoned SERPs against its own other SERPs, and each keyword weighted by its poisoned SERPs, so the comparison is not decided by which keywords are poisoned most.
Site table
| Site | Poisoned SERPs | Other SERPs | Change |
|---|---|---|---|
| facebook.com | 29.5% | 21.9% | +7.6 pts |
| instagram.com | 10.4% | 6.1% | +4.3 pts |
| youtube.com | 15.4% | 12.2% | +3.2 pts |
| quora.com | 5.0% | 2.4% | +2.6 pts |
| reddit.com | 23.0% | 20.7% | +2.3 pts |
| etsy.com | 2.8% | 0.7% | +2.1 pts |
| linkedin.com | 5.4% | 3.7% | +1.7 pts |
| mapquest.com | 3.4% | 2.5% | +0.9 pts |
| tiktok.com | 1.6% | 0.7% | +0.9 pts |
| justanswer.com | 2.0% | 1.4% | +0.6 pts |
| yelp.com | 10.8% | 11.5% | −0.7 pts |
Download the data (11 rows, one per site; gzipped CSV). It holds every number in the chart, with no account or keyword in it.
What each column means
- site
- The site, by registrable domain
- degraded_pct
- Share of poisoned SERPs since September 13 with the site in the top 10, each keyword weighted by its poisoned SERPs
- other_pct
- The same share on the same keywords’ other SERPs in the period, weighted the same way
On US keywords, 12.4% of poisoned SERPs have a foreign site, such as a .ca or .co.uk, in the top 10, against 6.2%.
Where it hit
More than four in five of the study’s datasets had at least one poisoned SERP. The map shows each country with at least 10 datasets and 10,000 judged SERPs: 45 of the study’s 53 countries.
Share of SERPs poisoned since Sep 13, by target country
- Under 5%
- 5–8%
- 8–10%
- 10–12%
- 12% or more
- Not enough data
A country needs 10 datasets and 10,000 judged SERPs to be coloured. Hong Kong and Singapore are too small to draw at this scale and show as a dot.
Country table
| Country | Poisoned SERPs | Judged SERPs |
|---|---|---|
| South Africa | 22.4% | 71,148 |
| Austria | 21.0% | 98,023 |
| United Arab Emirates | 17.1% | 51,995 |
| New Zealand | 16.8% | 202,522 |
| Switzerland | 16.1% | 85,365 |
| Belgium | 15.9% | 346,115 |
| Norway | 14.8% | 48,997 |
| Saudi Arabia | 14.8% | 51,850 |
| Thailand | 14.0% | 25,184 |
| Singapore | 13.7% | 78,115 |
| Ukraine | 13.0% | 17,763 |
| Australia | 12.9% | 532,709 |
| Colombia | 12.9% | 20,826 |
| India | 12.6% | 246,499 |
| Romania | 12.4% | 31,153 |
| Türkiye | 11.9% | 45,075 |
| Bulgaria | 11.8% | 29,440 |
| South Korea | 11.5% | 23,382 |
| Argentina | 11.3% | 60,130 |
| Sweden | 11.3% | 62,161 |
| Ireland | 11.1% | 93,587 |
| Brazil | 11.0% | 148,206 |
| Canada | 11.0% | 2,155,430 |
| Chile | 10.9% | 33,540 |
| Peru | 10.9% | 24,804 |
| Finland | 10.6% | 54,080 |
| Malaysia | 10.5% | 24,714 |
| Mexico | 10.4% | 227,640 |
| Poland | 10.2% | 81,224 |
| Hong Kong | 9.9% | 57,096 |
| Greece | 9.1% | 51,505 |
| United States | 8.8% | 18,383,115 |
| Indonesia | 8.3% | 27,225 |
| Taiwan | 8.0% | 29,960 |
| Portugal | 6.8% | 215,789 |
| Germany | 6.4% | 671,736 |
| Italy | 6.2% | 287,327 |
| Japan | 6.2% | 224,802 |
| Netherlands | 6.2% | 315,621 |
| Vietnam | 5.9% | 24,112 |
| Israel | 5.4% | 27,353 |
| France | 5.3% | 673,933 |
| Spain | 4.8% | 808,086 |
| Denmark | 4.6% | 16,977 |
| United Kingdom | 4.1% | 2,858,335 |
Download the data (45 rows, one per country; gzipped CSV). It holds every number on the map, with no account or keyword in it.
What each column means
- country
- Target country, ISO 3166 code
- judged
- SERPs whose keyword has usual sites: in its top 10 on at least 60% of five or more baseline SERPs, since September 13, 2026
- degraded
- Of those, poisoned SERPs: fewer than half of the usual sites on any page
Western Europe ran lowest, while its neighbors Austria, Switzerland and Belgium ran among the highest.
- “Near me” searches more. 13.6% of them, against 8.6% of everything else.
- The long tail more. 10.3% under 100 searches a month, 6.6 to 7.6% above that.
- Brand searches least. 5.3% for the most navigational phrases.
- Local targeting makes no difference. The same phrase tracked nationally and locally comes out the same, 8.2% and 8.4%.
Each ridge is one industry’s US SERPs, day by day; the grey ridge is every US SERP. Every industry peaked between September 27 and 29. What differs is how high.
Share of US SERPs poisoned per day, by industry
Every ridge is on one scale: the tallest, legal SERPs on Sep 27, is 54.6%. Grey: every judged US SERP. Shaded: the baseline. Dashed: Sep 13, when requests to Google started failing, and Sep 20, when poisoned SERPs first appear. The numbers are the share since Sep 13, or on the day you point at.
Industry table
| Industry | Since Sep 13 | Worst day | Judged SERPs |
|---|---|---|---|
| Legal | 17.4% | 54.6%, Sep 27 | 52,326 |
| Home services | 15.0% | 51.5%, Sep 29 | 220,014 |
| Weddings | 13.1% | 46.3%, Sep 28 | 30,353 |
| Transportation | 12.0% | 46.9%, Sep 29 | 315,335 |
| Events and entertainment | 11.3% | 43.6%, Sep 29 | 100,635 |
| Travel and hospitality | 11.0% | 41.2%, Sep 28 | 249,690 |
| Finance and banking | 9.7% | 35.8%, Sep 29 | 304,262 |
| Insurance | 9.7% | 36.2%, Sep 27 | 95,068 |
| Jobs and careers | 9.7% | 33.7%, Sep 29 | 300,309 |
| Education | 9.4% | 34.1%, Sep 28 | 58,543 |
| Software and technology | 9.2% | 29.7%, Sep 27 | 282,111 |
| All US SERPs | 8.8% | 31.2%, Sep 29 | 18,383,115 |
| Health and medical | 8.6% | 27.8%, Sep 29 | 257,653 |
| Telecom and internet | 8.6% | 34.4%, Sep 28 | 1,705,056 |
| Restaurants and food | 7.5% | 33.7%, Sep 29 | 270,868 |
| Beauty and fashion | 7.3% | 35.1%, Sep 28 | 104,490 |
| Home and garden | 7.3% | 31.3%, Sep 28 | 646,631 |
| Retail and shopping | 7.1% | 26.7%, Sep 28 | 2,374,360 |
| Real estate | 7.0% | 28.8%, Sep 27 | 5,341,198 |
| Automotive | 5.0% | 23.5%, Sep 28 | 587,745 |
| Pets | 4.4% | 13.8%, Sep 28 | 34,961 |
Download the data (1,323 rows, one per day and industry; gzipped CSV). It holds every number in the chart, with no account or keyword in it.
What each column means
- date
- Day the SERP was scheduled, UTC
- industry
- The keyword’s industry, or Unclassified
- judged
- US SERPs that day whose keyword has usual sites
- degraded
- Of those, poisoned SERPs: fewer than half of the usual sites on any page
Beyond the organic results
A poisoned SERP changes more than its organic results. For each part of the page, the chart compares a SERP with the keyword’s previous one: of what that part showed, how much it shows again.
What each part of the page keeps from the keyword’s previous SERP
- Baseline, Aug 20 to Sep 9
- Poisoned SERPs
Keeps less on poisoned SERPs
- Organic results88.1% → 22.3%
- People also ask87.1% → 33.7%
- People also search for87.6% → 34.2%
- AI Overview sources68.1% → 37.5%
- Local pack85.8% → 66.5%
- Image pack85.2% → 70.2%
- Video79.6% → 65.5%
- Discussions and forums90.6% → 77.1%
- Knowledge panel87.7% → 76.5%
Keeps more on poisoned SERPs
- Shopping93.0% → 93.1%
- Ads70.2% → 83.9%
Of the sites, questions, searches or businesses a part showed on the keyword’s previous SERP, the share it shows again, where both SERPs show that part. Each keyword’s poisoned SERPs against its own baseline, each keyword weighted by its poisoned SERPs.
Part table
| Part | Kept from the previous SERP | Shown on | ||||
|---|---|---|---|---|---|---|
| Baseline | Poisoned | Others since Sep 13 | Baseline | Poisoned | Others since Sep 13 | |
| Organic results | 88.1% | 22.3% | 77.3% | 100.0% | 100.0% | 100.0% |
| People also ask | 87.1% | 33.7% | 72.7% | 87.1% | 77.1% | 84.7% |
| People also search for | 87.6% | 34.2% | 79.4% | 85.9% | 61.0% | 83.3% |
| AI Overview sources | 68.1% | 37.5% | 62.5% | 53.6% | 48.2% | 60.8% |
| Local pack | 85.8% | 66.5% | 70.3% | 39.3% | 41.6% | 40.7% |
| Image pack | 85.2% | 70.2% | 81.8% | 22.5% | 33.8% | 23.2% |
| Video | 79.6% | 65.5% | 77.8% | 21.8% | 19.6% | 23.1% |
| Discussions and forums | 90.6% | 77.1% | 90.8% | 19.2% | 22.4% | 19.7% |
| Knowledge panel | 87.7% | 76.5% | 81.6% | 4.2% | 5.3% | 5.4% |
| Shopping | 93.0% | 93.1% | 95.1% | 13.7% | 13.8% | 14.2% |
| Ads | 70.2% | 83.9% | 72.6% | 9.5% | 8.6% | 7.4% |
Download the data (11 rows, one per part of the page; gzipped CSV). It holds every number in the chart and the table, with no account or keyword in it.
What each column means
- part
- The part of the page
- kept_baseline_pct
- Of the sites, questions, searches or businesses the part showed on the keyword’s previous SERP, the share it shows again, where both SERPs show the part: baseline SERPs, August 20 to September 9
- kept_degraded_pct
- The same for poisoned SERPs since September 13, 2026, each against the keyword’s previous SERP
- kept_other_pct
- The same for the same keywords’ other SERPs since September 13, 2026
- shown_baseline_pct
- Share of baseline SERPs showing the part
- shown_degraded_pct
- Share of poisoned SERPs showing the part
- shown_other_pct
- Share of the same keywords’ other SERPs since September 13, 2026 showing the part
Only the parts advertisers pay for, shopping and ads, hold steady.
What it does to tracked sites
Where a project’s own site is one of the keyword’s usual sites, it was out of the top 10 on 60.8% of poisoned SERPs, against 2.7% in the baseline. Then it passes: four in five spells of poisoning are a single SERP, 67.9% of poisoned SERPs have the usual sites back within three days, and 92.5% within two weeks. A tracked site that fell out of the top 10 for a day in late September and came straight back was most likely poisoned.
How to tell it from a real ranking change
- The page is complete. No error, no captcha, the same template. Nothing in the HTML marks it.
- The results don’t match the keyword’s history. Compare with its usual sites over the previous weeks, not with yesterday, which may have been poisoned too.
- It doesn’t last. Four in five spells are a single SERP. A real ranking change stays.
- Check page 2. It is poisoned one and a half times as often, mostly under a page 1 that looks fine, and it is where Google most often ignores part of the query.
- Search Console disagrees. If a site “disappears” and Search Console still shows impressions and clicks, nothing happened to the site.
Method
- SERPs. The study draws on a representative sample of the SERPs Nozzle tracks for its customers, from August 20 to October 21, 2026: 53 countries, desktop, iPhone and Android, and targeting from whole countries down to postal codes.
- Judged. A keyword’s usual sites are the sites in its top 10 on at least 60% of its SERPs from August 20 to September 9, the baseline, with five SERPs or more. Only those keywords’ SERPs with organic results are judged, and only judged SERPs are counted: 56.1 million SERPs, 29.9 million of them since September 13.
- Poisoned. Fewer than half of the keyword’s usual sites on any page of the SERP. The test needs no later SERP, so a day’s figure is final when it is published. Back means at least 80% of the usual sites on a later SERP of the keyword.
- Page 2. The same test against the keyword’s usual page-2 sites, on the desktop SERPs judged on both pages: 4.7 million since September 13. A SERP poisoned on both pages counts once. A page ignores part of the query when a word of the query that at least 60% of the keyword’s baseline results carried is on 20% of them or fewer; on page 1 such pages stay in the poisoned count when they also fail the usual-sites test, 0.9 of page 1’s 8.7 points.
- Comparisons. Each keyword is compared with itself and weighed by its poisoned SERPs. What a part of the page keeps is the share of its items on the keyword’s previous SERP, at most 48 hours earlier, that it shows again.
- Images. Nozzle’s stored page-1 HTML, rendered in Chrome with Google’s scripts blocked. The sites shown are Google’s results, not Nozzle customers.