What Is a Poisoned SERP? Should You Care?

5/5 - (5 votes)

A poisoned SERP is a full Google results page that looks normal but has most of the usual sites for that keyword swapped out. Google serves it to automated requests, such as rank trackers, SERP APIs and AI scrapers. The page loads with an HTTP 200 code and no CAPTCHA, so the tool saves it as a real ranking.

The usual symptom is a client keyword that falls from position 3 to “not in top 100” overnight while Search Console shows nothing unusual, then climbs back two days later. Rank trackers have been picking up these pages since late September 2026, and a small share still turns up today.

What a Poisoned SERP Is and How It Looks

Take the search “iphone” in the US, in English. A real person sees Apple first, then Wikipedia and other familiar pages. A poisoned SERP for the same request can lead with YouTube videos and other pages that are still about iPhones, so nothing about it looks broken.

The layout, the SERP features and the topic all check out. However, the list of sites is wrong. DataForSEO counts a SERP as poisoned when fewer than half of the sites that usually rank for the keyword appear anywhere on it.

By contrast, Nozzle’s test is tighter. A reading counts as poisoned when fewer than half of the usual sites appear and those sites come back on a later fetch. A site only counts as “usual” if it sat in the top 10 on at least 60% of past readings.

Some poisoned pages are harder to catch. In one Semrush example, a search in Brazil for a baby stroller model returned a dental supply page, because both products share the Portuguese word “carrinho.” Other pages open with the right results and then drift into loosely related ones further down.

Poisoned SERP vs. a Real Drop and SEO Poisoning

A real ranking drop shows up for people too, so Search Console clicks and impressions fall with it. A poisoned SERP only reaches bots, which is why your traffic stays flat while the tracker panics. SEO poisoning is an older and separate problem, where attackers push malware or phishing pages into the results real people see. For that side of search spam, our breakdown of the September 2026 Google spam update covers what Google went after this fall.

Poisoned SERP Timeline, September to November 2026

Google had been making scraping harder for a long time before this fall. First, in January last year, it began to require JavaScript for search and added a bot check called SearchGuard. Then, in September last year, it dropped the num=100 setting that let tools pull 100 results in one request.

In the fall of 2026, requests failed first. The wrong pages arrived a week later, and those were much harder to catch.

Date (2026)What happened
September 13Requests to Google start failing in large numbers in Nozzle data.
September 20The first poisoned SERPs appear among requests that get through.
September 24Google starts the September spam update. AccuRanker sees swapped pages late that day.
September 26Poisoned pages hit DataForSEO, at about 20% to 30% of scraped results.
September 27 to 29The peak. About a quarter of all SERPs in the Nozzle sample are poisoned.
October 1 onwardThe daily rate settles at about 5% to 11% in the Nozzle sample.
October 6 and 7DemandSphere adds error and sampling rates next to its rank data and puts spurious results at about 11% of daily SERPs.
October 8The spam update finishes. Poisoned SERPs keep coming.
October 10Nozzle publishes its study. Google starts serving fewer poisoned pages to DataForSEO.
October 16SE Ranking posts on X that Google is returning incomplete or unreliable results to rank trackers across the industry.
October 22Ahrefs puts the poisoned share at 5% to 20% of results, depending on the data provider.
October 27Semrush confirms it has seen the issue since late summer and adds checks.
November 16DataForSEO ships the resolve_poisoned_results fix.

Google has made no public statement about poisoned SERPs or what causes them. Meanwhile, Nozzle ties the change to AI scraping and grounding, where bots pull search results to feed AI answers. DemandSphere CEO Ray Grieselhuber sees Google trying to keep AI labs from using its results for live retrieval, with rank trackers caught in the middle.

How Many Google Results Were Poisoned

Nozzle has the largest public sample, 27.5 million SERPs from 49 countries between September 13 and October 9. Of those, 8% were poisoned. By contrast, only 0.17% of SERPs were poisoned before the problem began.

That average hides a sharp spike. The daily rate climbed from under 2% on September 20 to 27% on September 29, then fell back into a band of about 5% to 11%.

Line chart of the daily share of poisoned Google SERPs in the Nozzle sample, rising from 1.9 percent on September 20 to a peak of 27 percent on September 29, then falling to 4.2 percent by October 9, against a 0.17 percent baseline

A poisoned page also has a different shape. Only 20.7% of its sites carry over from the last reading, against 88.5% on a normal page, and 58.6% of poisoned pages have lost tracked sites from the top 10, against 2.7% at baseline. AI Overviews show up less often, on 34.9% of poisoned pages compared with 54.2% normally, while image packs rise to 44.2% from 23.0%.

Social sites also gain ground on these pages. Reddit went from 20.8% to 28.8% of top 10 appearances, and Facebook from 28.3% to 33.8%.

Most poisoned readings do not last. Four in five lasted only one fetch, and the usual sites came back within three days in 77.9% of cases, with a median of two days. Meanwhile, ads barely changed at all.

DataForSEO took a heavier hit at first, with poisoned pages making up about 20% to 30% of its scraped results from September 26. Two retries checked against past results still left 2% to 3% of pages poisoned. Google has served it fewer poisoned pages since October 10, but the share has not reached zero.

Keywords and Countries Hit Hardest by Poisoned SERPs

Local and low-volume searches took the worst of it. “Near me” searches were poisoned 11.4% of the time, against 7.8% for everything else, and long-tail phrases under 100 monthly searches came in at 9.4%. By contrast, branded searches had the lowest rate, at 5.3%.

In the US, legal keywords topped the industry list at 15.9%, with home services next at 12.7% and pet keywords lowest at 4.3%. An agency that runs SEO for law firms or plumbers should expect more false alarms than one that works for an ecommerce pet brand.

SegmentPoisoned rate (Sept 13 to Oct 9)
Austria18.8% (highest country)
Australia11.5%
India10.9%
Canada10.5%
UK3.0% (lowest country)
Postal code SERPs10.7%
City SERPs8.4%
Country-wide SERPs7.8%
iPhone8.3%
Android7.1%

On the peak day, September 29, some markets were far worse than the averages suggest. Canada hit 48.1% and Australia 47.5%, the US reached 29.4%, and the UK, by contrast, stayed at 2.8%.

How to Spot a Poisoned SERP in Your Rank Tracker

Most cases can be caught without special software. So before you tell anyone about a sudden drop, run it through these checks.

  1. Pull the full top 10 for the keyword, not just your own position, and compare it with the last three to four weeks. If fewer than half of the usual domains are there, the reading is suspect.
  2. See who else moved. A real drop usually hits your page alone or a handful of pages, while a poisoned page swaps out most of page 1 at once.
  3. Wait a day, or fetch again from another location or route. A drop that reverses within two to three days points to poisoning.
  4. Look at the shape of the page. Missing AI Overviews, extra image packs and results that stop by page 4 are common on poisoned readings.
  5. If you use DataForSEO, open the check_url link in an incognito window. A normal browser almost never gets a poisoned page.
  6. Steady impressions and clicks in Search Console for that query mean the real ranking did not fall.
  7. Before you blame the tracker, rule out a Google update, a site change or lost backlinks.

Step 2 is the quickest filter, because a real Google update rarely clears Apple and Wikipedia off an “iphone” result on the same morning it moves your client.

Handling Poisoned SERP Data in Client Reports

The costliest outcome is a panicked client call about a drop that never happened. However, a few reporting rules prevent most of them.

SEO consultant walking a client through a rank tracking report
Photo: Pexels

First, set rank alerts to fire only when two readings in a row show the same change, and treat any one-day drop as unconfirmed. For volatile keywords, report the position that held for three days.

Annotate every rank chart from September 20, 2026 onward. Data from September 26 to mid-October needs the most care, because the September spam update ran from September 24 to October 8 and real moves and fake ones overlap in that window.

Treat Search Console as the record. It counts impressions and clicks from real people, and poisoned pages never reach those people. Export the data now as well, because Search Console only keeps 16 months. If you build client dashboards, our list of AI SEO reporting tools covers tools that pull Search Console data in next to rankings.

Tell clients early and in plain words. A line like this works: “Google sends some SEO tools fake results pages, so we confirm every drop in Search Console before we act on it.”

How Rank Trackers and SERP APIs Responded to Poisoned SERPs

Google controls what it sends, so no vendor can stop poisoned pages outright, and each one took a different route around them.

DataForSEO went the furthest. Its resolve_poisoned_results option makes one extra request to the “Web” version of the same search, compares the organic results and swaps in the clean ones, marked from_web. Each task then costs one more SERP page, and the option cannot repair data from late September to mid-October.

At Semrush, obviously invalid pages get dropped during collection, some data sources have changed and suspect results go back for a second collection. Ahrefs added filters that reject abnormal results and request new ones, then cleaned up past results that may have been hit.

One in every 200 tracked keywords at AccuRanker gets a quality check. Its flagged share then fell from 30.2% of desktop pages on September 25 to about 1.5% from October 1 to 9.

SISTRIX turned off on-demand updates, narrowed its focus to the top 10 and now publishes only results that pass its quality checks. DemandSphere took the opposite tack and made the problem visible, with an Observability feature that puts error and sampling rates next to ranking data.

So if your tracker has been silent, ask support what share of readings they reject and what they did with the data from late September. For local rank checks, our guide to local Google Business rank tracking tools can help you compare options.

How Poisoned SERPs Affect Each Type of SEO User

Your real rankings are not at risk. However, the data you read about them is, and how much that matters comes down to what you do with that data.

In-House SEOs

Only a little, unless you work in legal, home services or local search. So, with one boss to report to and your own Search Console access, a note on the September charts and the two-reading rule cover most of it.

Agencies

Agencies carry the most risk from a single bad reading, because one false drop in a client report can cost an account’s trust. Audit your September and October reports, warn clients and ask your tracker vendor how it filters bad pages. Check hardest on clients in Canada, Australia or Austria.

SEO Tool Builders

At DataForSEO volumes, retries alone still left 2% to 3% of pages poisoned. If you build SEO tools, a poisoned page goes into your database as real data unless you catch it, so you need history checks, a second source and a plan for the next time Google changes the rules.

Site Owners Who Only Use Search Console

You can ignore this one. Search Console reflects what real people saw and clicked, so poisoned pages never touch it. If a freelancer sends you a scary rank drop, ask for the Search Console clicks for the same dates.

Common Questions About Poisoned SERPs

Do AI Overview and local pack trackers get poisoned too?

Yes. Organic results take the biggest hit, but AI Overviews, People Also Ask boxes and local packs can also come back wrong on a poisoned page. Read AI visibility numbers from the same weeks with the same care.

Why do some trackers report much lower rates than others?

How a tool collects data changes what Google sends it. From September 30 to October 9, one Nozzle collection route got 1% to 7% poisoned pages while another got 9% to 30%, so a 1% rate and a 20% rate can both be honest.

Did the September spam update cause poisoned SERPs?

The timing overlaps, but the update alone does not explain them. Poisoned pages first appeared on September 20, four days before the update began, and they kept coming after it ended on October 8.

Will Google stop serving poisoned SERPs?

Nobody outside Google knows. Rates have dropped a lot since early October, yet AccuRanker saw Google tighten detection again on October 15, mostly on mobile. So plan on poisoned pages staying a regular part of rank tracking.

Add Comment