Back to blog
seo news11 min read

Poisoned SERPs: Google Is Feeding SEO Tools Swapped Results. Here's What to Trust

Since September 2026 Google has served fake-looking results pages to rank trackers. What the vendors measured, which data it reaches, and which it can't.

Aditi ChaturvediNovember 16, 2026
TL;DR

Since September 2026, Google has been answering some automated search requests with poisoned SERPs: complete, normal-looking results pages whose sites have been swapped out. Rank trackers and SERP data providers record them as real. Google has not commented.

  • What it looks like: A full results page with the usual layout, where fewer than half of the sites that normally rank for the keyword appear. Nothing on the page signals that it is wrong.
  • How much: Ahrefs puts it at 5% to 20% of results depending on the data provider. Nozzle measured a poisoned page 1 on 15.0% of SERPs since 13 September, against 0.17% before, with a one-day high of 68.6% for page 1 or 2 on 2 November.
  • Who is affected: Ahrefs, Semrush, DataForSEO, SerpApi and Nozzle have all confirmed it in their own data. Anything built from automated Google queries is exposed: positions, share of voice, volatility indexes.
  • What it does not reach: Search Console, GA4 and a crawl of your own site. None of them is collected by querying Google's results pages, so this mechanism cannot touch them.
  • What it is not: Proof that all third-party SEO data is corrupt. Most tracked results are still correct, vendors are filtering, and competitor data still has no first-party substitute.

Run your scoreboard on your own data. Treat scraped rankings as a sample to verify, not a verdict.

This is a news report on an unconfirmed but widely measured change, with figures as published by the data providers up to 16 November 2026. CrawlRaven works from data Google and your own site provide directly: it ranks content opportunities from Search Console evidence and saved priorities, and runs a separate technical audit, from $29/month. Try CrawlRaven free for 7 days: 1 website, no credit card →

Imagine your rank tracker shows a client's money page falling out of the top 10 overnight. You open Search Console and the clicks are exactly where they were yesterday. Since September 2026, that mismatch has a name.

Google is answering some automated search requests with poisoned SERPs: complete, normal-looking results pages whose sites have been swapped out. Rank trackers store them as real. Ahrefs, Semrush, DataForSEO, SerpApi and Nozzle have all confirmed it. Google has not commented.

This report covers:

  • What a poisoned SERP is, and how the vendors detect one.
  • The timeline from mid-September to mid-November.
  • How much data each vendor says is affected.
  • Whether this means all third-party SEO data is now corrupted (it does not).
  • Which of your data sources a poisoned page can reach, and what to do this week.

What a poisoned SERP is

A poisoned SERP is a full results page, with the usual layout and features, where most of the organic results are not the ones that normally rank. DataForSEO's description is the clearest: nothing on the page directly signals that anything is off.

Poisoned SERP: in plain English

Nozzle calls a page poisoned when fewer than half of a keyword's usual sites appear anywhere on it. “Usual” means sites that were in the top 10 on at least 60% of that keyword's results between 20 August and 9 September 2026. DataForSEO uses the same fewer-than-half test.

Semrush describes two forms:

  • Clearly off. The whole page is wrong, with results that do not match the query even on page one.
  • Subtly off. The top looks right, but lower results only loosely match. Its example is a Brazilian search for a stroller model that returned a dental supply store, because both use the word “carrinho”.

What fills the gap matters too. On poisoned pages, Nozzle found organic results kept only 30.0% of the previous page's sites, against 88.2% normally. Facebook and YouTube took a larger share of the top 10.

How it unfolded

The dates below come from the vendors' own publications. All are 2026.

  • 13 September: Nozzle sees automated requests to Google failing in large numbers.
  • 20 September: the first poisoned pages appear in Nozzle's data. DataForSEO dates its own onset to 26 September.
  • 30 September: a first peak, with 43.5% of the SERPs Nozzle judged poisoned on page 1 or 2.
  • Mid October: an easing. DataForSEO reports fewer poisoned pages from 10 October.
  • 22 to 27 October: Ahrefs and then Semrush publish statements confirming the problem.
  • 2 November: the worst day so far, at 68.6% for page 1 or 2.
  • 6 November: Ahrefs says its page one data is largely recovered and stable.
  • 12 November: SerpApi says it still sees the pattern intermittently, in Google Search and Google Shopping.
Timeline

Poisoned SERPs, September to November 2026

As measured and reported by the data providers themselves

20 Aug – 9 SepNormal
Nozzle's baseline period

Before the change, 0.28% of the SERPs Nozzle judged were poisoned on page 1 or page 2. That is the noise floor.

13 SepOnset
Automated requests start failing

Nozzle records requests to Google failing in large numbers. It uses this date as the start of its measurement window.

20 SepOnset
First poisoned pages appear

Nozzle's data shows the first complete, normal-looking results pages whose sites have been swapped out. DataForSEO dates its own onset to 26 September.

30 SepPeak
First peak

43.5% of the SERPs Nozzle judged on both pages are poisoned on page 1 or page 2.

Mid OctEasing
A temporary easing

DataForSEO reports noticeably fewer poisoned SERPs from 10 October. Nozzle's rate bottoms out around 14 and 15 October.

22 – 27 OctConfirmed by vendors
Ahrefs and Semrush go on record

Ahrefs puts the share of affected results at 5% to 20% depending on the data provider. Semrush says it now excludes clearly invalid results. Google has not commented.

2 NovPeak
The worst day so far

68.6% of the SERPs Nozzle judged on both pages are poisoned on page 1 or page 2. Page 1 alone peaked earlier, at 34.8% on 27 October.

6 NovPartial recovery
Ahrefs: page one largely recovered

Ahrefs says its page one data is stable again, but it is rejecting more page two results, which can delay updates for some keywords.

12 NovOngoing
SerpApi: still intermittent

SerpApi reports the pattern in its Google Search and Google Shopping APIs, says it still sees it intermittently, and says it has seen similar behaviour on Bing.

Dates from Nozzle, DataForSEO, Ahrefs, Semrush and SerpApi, checked 16 November 2026. All dates are 2026.

The timing overlapped with Google's September spam update, which made early drops hard to read. DataForSEO advises separating the two when you review that period. The simplest test: a real spam-update loss also shows up in Search Console, and a poisoned page does not.

How much data is affected

Every vendor counts differently, so the headline numbers disagree. They are not contradictory. They measure different things.

  • Nozzle (updated 3 November): a poisoned page 1 on 15.0% of SERPs since 13 September, against 0.17% before. Page 1 or 2 was poisoned on 27.6%. The sample is 68.7 million SERPs across 58 countries on desktop, iPhone and Android.
  • Ahrefs (22 October): between 5% and 20% of search results, depending on the data provider. It published no method or sample.
  • DataForSEO (23 October): about 20% to 30% of all scraped results since 26 September. Two retries with historical validation still left 2% to 3% poisoned.
  • Semrush (27 October) and SerpApi (12 November): no figures, but both confirm the pattern and describe their filtering.
The numbers

How much is poisoned? Five vendors, five yardsticks

Each measures something different, so the figures do not add up

NozzleUpdated 3 Nov
15.0%

Share of SERPs with a poisoned page 1 since 13 September, against 0.17% before it.

68.7 million SERPs across 58 countries. The one-day high for page 1 or 2 was 68.6%.

Ahrefs22 Oct
5–20%

Share of search results affected, depending on which data provider supplied them.

No method or sample size published. On 6 November it said page one had largely recovered.

DataForSEO23 Oct
20–30%

Share of all scraped results that were poisoned, since 26 September.

Two retries with historical checks still left 2–3% poisoned. Fewer since 10 October.

Semrush27 Oct
No figure

Confirms two forms: pages clearly off from page one, and pages only loosely matching lower down.

Says it excludes clearly invalid results and re-collects flagged ones.

SerpApi12 Nov
No figure

Sees it in its Google Search and Google Shopping APIs, on page one and beyond.

Still intermittent after its fixes. Reports similar behaviour on Bing, without detail.

Vendor-published figures, 2026, checked 16 November. None is an independent audit, and Google has confirmed none of them.

Two Nozzle findings matter more than the headline rate. A tracked site was out of the top 10 on 63.7% of poisoned SERPs, against 2.6% normally. So when a poisoned page gets through, it usually shows up as a dramatic, false drop.

Most episodes are also short. Nozzle found 45.1% of poisoned SERPs had their usual sites back within three days and 89.0% within two weeks. That is why a one-day drop deserves suspicion before a client email.

Why Google might be doing it

Google has not confirmed the behaviour, its trigger or its target. What is on the record:

  • Google's spam policies already name “scraping results for rank-checking purposes” as machine-generated traffic that violates its policies and terms of service.
  • Google is in court with SerpApi, a large SERP data reseller, over scraping. Our report on the Google SerpApi lawsuit covers the case and the access restrictions that came before it.
  • Semrush, Ahrefs and others cite industry discussion that the target is large-scale data collection for AI systems, with SEO tools caught in the same net.

Nozzle makes one point worth repeating. Google is not refusing these requests, which a scraper would notice and route around. It answers them with plausible wrong pages, which costs the scraper more to detect than a block would.

Is all third-party SEO data corrupted now?

No. That claim is circulating, and it overshoots the evidence. Here is what the evidence supports:

  • The mechanism is narrow. It reaches data collected by sending automated queries to Google results pages. A backlink index, built by crawling the web, is a different pipeline entirely.
  • Most results are still correct. Even the highest sustained vendor estimate leaves most tracked results accurate, and the worst days were spikes, not the norm.
  • The big vendors are filtering. Ahrefs says it rejects abnormal results and is cleaning past ones. Semrush says it excludes clearly invalid pages and re-collects flagged ones.
  • Competitor data has no first-party substitute. Your own data cannot tell you where a rival ranks.

The accurate version is sharper and more useful: scraped Google rankings are now a sample that needs verifying, and your own first-party data is the thing you verify them against.

Which data a poisoned SERP can reach

One question sorts every source: was this number produced by querying Google's results pages? If yes, it is exposed. If no, this mechanism cannot touch it.

Exposure check

Which of your data can a poisoned SERP reach?

The test is simple: was it collected by querying Google?

Rank tracker positions
Automated Google queries

A swapped page reads as a ranking drop. Nozzle found tracked sites out of the top 10 on 63.7% of poisoned SERPs.

Exposed
Competitor visibility and share of voice
Built from tracked SERPs

Inherits every bad snapshot, for every domain on the page at once.

Exposed
SERP volatility indexes
Day-over-day SERP changes

A swapped page looks exactly like volatility. Read spikes since September with care.

Exposed
Keyword tools' SERP views
Stored SERP snapshots

Only the parts read from the results page. Vendors say they are filtering and cleaning history.

Indirectly
Search Console
Google's logs of real searches

Not collected by querying Google, so this mechanism cannot reach it. It still has its own bugs.

Not exposed
GA4 and server logs
Your own site's visits

Records people who arrived, not pages a bot was shown.

Not exposed
A crawl of your own site
Fetches your pages

Never touches a Google results page.

Not exposed

“Not exposed” means out of reach of this mechanism, not free of error. Search Console's Generative AI report under-logged impressions on 9 to 11 November 2026.

  • Exposed: rank tracker positions, competitor visibility and share-of-voice scores, and SERP volatility indexes, which read a swapped page as real movement.
  • Indirectly exposed: the parts of keyword tools that show a stored results page for a keyword.
  • Not exposed: Search Console, which reports Google's own record of real searches; GA4 and server logs, which record actual visits; and a crawl of your own site.

That is why Search Console is the natural tie-breaker. Its average position is calculated across real impressions on real devices, not from one snapshot a datacenter browser was shown.

Your own data is not error-free either

“Trust only your own data” is close to right, but not quite. First-party data is immune to this problem, not to every problem:

  • It has bugs. Google confirmed a logging error that lowered impressions in the Search Console Generative AI performance report from 9 to 11 November 2026. It affected reporting only, not rankings.
  • It forgets. Search Console keeps 16 months by default. Our guide to Search Console data retention covers how to keep more.
  • It only sees you. No competitor rankings, and no results page you did not appear on.

What to do now

None of this needs Google to confirm anything first.

  1. Annotate from 20 September. Mark the period in every dashboard and report that uses scraped rankings, so nobody misreads it later. Our guide to SEO annotations covers how.
  2. Confirm drops against Search Console before acting. A ranking drop with flat impressions and clicks for that page and query is a data problem, not a ranking problem.
  3. Require two consecutive readings. DataForSEO recommends alerts fire only when two checks in a row agree, and treats one-day drops as unconfirmed.
  4. Start the BigQuery export today. It only collects from the day you switch it on. The BigQuery export guide walks through setup.
  5. Ask your vendors two questions. How do you detect poisoned pages, and have you corrected historical data since September?
  6. Report your own performance from your own data. Clicks and position from Search Console, sessions from GA4, with scraped competitor data clearly labelled as such. The GSC regex generator splits queries into branded and non-branded segments for that report.
Opinion· Aditi's take

This episode should end a bad habit: treating a rank tracker screenshot as a verdict. Your scoreboard should be the data Google records about real searches of your site. Rank trackers are still useful for reconnaissance: who moved, which competitor appeared, what the results page looks like. They are reconnaissance, not the scoreboard.

Key Takeaways

  • →What changed: Since September 2026, Google has served some automated requests complete, normal-looking results pages with the usual sites swapped out. Google has not commented.
  • →How big: Vendor figures range from 5% to 20% (Ahrefs) to a 15.0% page 1 rate since 13 September (Nozzle), with spikes far higher on single days.
  • →What it reaches: Anything built from automated Google queries: rank positions, share of voice, volatility indexes. Not Search Console, GA4 or your own crawl.
  • →What to do: Annotate from 20 September, confirm every drop against Search Console, require two consecutive readings, and report your own performance from first-party data.

Sources used in this post

Frequently asked questions

What are poisoned SERPs?

Poisoned SERPs are Google results pages that look complete and normal but contain a different set of sites from the ones that usually rank for the query. Since September 2026, rank trackers and SERP data providers have reported receiving them in response to automated requests. Nozzle defines a page as poisoned when fewer than half of the keyword's usual sites appear anywhere on it.

Is Google deliberately poisoning search results?

Google has not said. As of 16 November 2026 it has made no public statement on the reports. Ahrefs, Semrush, DataForSEO, SerpApi and Nozzle have all confirmed the pattern in their own data. The common reading is that it targets large-scale scraping, especially for AI systems, and Google's spam policies already name scraping results for rank-checking as a violation. That motive is an inference, not a confirmed fact.

How much rank tracking data is affected?

Estimates range widely because vendors measure different things. Ahrefs said 5% to 20% of results depending on the data provider. DataForSEO said 20% to 30% of its scraped results since 26 September, falling to 2% to 3% after retries. Nozzle measured a poisoned page 1 on 15.0% of SERPs since 13 September, with a one-day high of 68.6% for page 1 or 2 on 2 November.

Is Google Search Console data affected by poisoned SERPs?

No evidence says it is. Search Console reports Google's own record of real searches that showed your site, delivered directly to verified owners, so it is not collected by querying results pages. Nozzle suggests using it as the check: if a tracker says a page vanished but Search Console still shows its impressions and clicks, the page probably did not change. Search Console does have separate bugs, such as the 9 to 11 November 2026 logging error.

Is all third-party SEO data corrupted now?

No. The poisoning affects data collected by sending automated queries to Google, mainly rankings and anything built from them. Most tracked results are still accurate, the major vendors are filtering bad pages out and cleaning history, and data that is not scraped from Google, such as a backlink index, is a different pipeline. The accurate statement is that scraped Google rankings now need verifying.

How can I tell if a ranking drop is real or a poisoned SERP?

Check three things. First, does Search Console show a matching fall in impressions and clicks for that page and query? Second, does the drop persist across two or more consecutive checks? Third, did the whole results page change, with most usual competitors also gone? A one-day drop that Search Console does not confirm, on a page where everyone else also vanished, is most likely poisoned.

When did poisoned SERPs start?

Nozzle's data shows automated requests failing in large numbers from 13 September 2026 and the first poisoned pages on 20 September. DataForSEO dates its onset to 26 September. Rates eased in mid-October, then rose again to the highest level so far on 2 November. On 6 November Ahrefs said its page one data had largely recovered.

Does this affect Bing or other search engines?

SerpApi said on 12 November 2026 that it has also observed the behaviour on Bing, but it gave no figures or detail. Every other published measurement covers Google only.

Aditi Chaturvedi
About the Author

Aditi Chaturvedi

15+ years of growing SaaS websites through SEO | Author, 200-Point Audit Checklist

Aditi has spent 15+ years helping SaaS companies scale organic traffic through technical SEO and content strategy. She is the author of the CrawlRaven 200-Point Audit checklist used by agencies and in-house teams to systematically improve search performance.

poisoned serpsgoogle data poisoningrank trackingserp scrapingrank tracker accuracysearch console datafirst-party dataseo news

Poisoned SERPs reach data collected by querying Google. They do not reach Search Console, GA4 or a crawl of your own site.

Run your reporting on data nobody can swap out

CrawlRaven ranks content opportunities from Search Console evidence and saved priorities, with GA4 where it qualifies, and runs a separate technical audit. Start with a 7-day free trial, no credit card; plans run from $29/month. See the pricing page for current plans.

A rank tracker reports what a bot was shown. Search Console reports what real searchers saw. Since September, only one of those has been reliable on every day.

CrawlRaven connects Google Search Console and GA4, runs technical SEO checks, and joins all three into one prioritized fix list, so you know what is broken, what it is costing you, and what to fix first.

✓ No credit card required·Technical SEO audit·GSC + GA4 + full-site crawl
7-day free trial, no credit card

Stop exporting. Start shipping.

Connect Search Console, import your Ahrefs or Semrush lists, and get content priorities and separate audits. Start a 7-day free trial. Plans from $29/month.

1
Content priority view
1
Separate audit workspace
CSV
Findings export