Poisoned SERPs: Your Rank Tracker May Be Lying, and Google Isn’t Saying Why

Picture opening your rank tracker on a Monday morning. A keyword you have held for two years has vanished from page one. Search Console shows steady impressions, and the pipeline hasn’t moved. Two days later, the ranking is back.

That pattern may not be your SEO failing. Since mid-September, rank-tracking vendors have reported that Google is returning results pages that look completely normal but swap out most of a keyword’s usual sites, and the tools save them as valid data. Nozzle, which published the largest study so far, calls these “poisoned” SERPs.

This article separates what the data proves from what vendors are guessing, explains why “Google is poisoning search results” is both accurate and easy to misread, and ends with what B2B teams should do before the next dashboard panic. It’s written for CMOs, marketing leaders and founders who report on organic performance.

What happened, in order

The sequence below uses dates reported by Nozzle and the trade press. Several events overlap, which is part of why the picture is murky.

The scale is large. Nozzle’s study covers 63.1 million SERPs collected from August 20 to October 9, on desktop, iPhone and Android, with targeting from whole countries down to postal codes. About four in five of its datasets had at least one poisoned SERP. The company sells rank tracking, so this is vendor research, not an independent audit.

What poisoning actually means

A poisoned SERP is a complete, normal-looking Google results page where most of the sites a keyword usually shows have been replaced. There is no error and no captcha. Nozzle counts a page as poisoned when fewer than half of the keyword’s usual sites appear on it (sites in the top 10 on at least 60% of that keyword’s SERPs between August 20 and September 9) and most of them return on a later check.

The distinction that matters is what a collection tool does next. A refused request tells a scraper to try again. A poisoned one looks like success, so it gets stored. That is why it quietly corrupts reports.

The size of the distortion is stark. On poisoned pages, only 20.7% of the previous SERP’s organic sites remained, compared with 88.5% normally. For keywords where a tracked site was a usual top-10 result, it was missing from the top 10 on 58.6% of poisoned pages, against 2.7% at baseline.

What it does not mean: nothing published so far shows that real searchers are being served these swapped pages. The data comes from automated requests. Nozzle’s own guidance is to check Search Console, because if clicks and impressions for the query continue, your site probably did not disappear.

Why it happened: what is proven and what is not

Google has not confirmed poisoning or given a motive, so every explanation below sits on a spectrum from measured to speculative.

What Google has said on the record is about adjacent actions. In December 2025 it sued the scraping company SerpApi, and Google’s own announcement accuses it of circumventing security measures and overriding site directives. Those are Google’s allegations, and the post does not carry SerpApi’s response. On the goto redirects, Google said it has a long history of deploying technical measures against evolving forms of abuse. Neither statement addresses poisoned results.

Practitioners are cautious. Per Search Engine Roundtable, Derek Perkins of Nozzle describes three tactics: wide IP blocking of residential proxy networks, more captchas, and injecting bad data into suspect SERPs. Barry Schwartz calls the last one the most concerning because tools may not spot it quickly. Ray Grieselhuber’s view, reported by PPC Land, is that Google doesn’t care about tools in the AI search and SEO space, and that they are caught in the crossfire. Lily Ray said rank trackers are too unreliable at the moment, and Glenn Gabe noted that Ahrefs showed a drop that Semrush reported late and Sistrix did not show at all.

Where AI content and SEO spam fit

This is the part most coverage blurs, so precision helps. The evidence ties poisoning to AI scraping, meaning bots that pull Google results to feed AI answers. It does not tie poisoning to AI-written content or to SEO spam.

• The AI-scraping link. Grieselhuber argues Google wants to stop AI labs from using its SERPs as a live retrieval layer. If so, any AI agent that pulls SERP data is exposed to the same bad pages, a risk Ryan Doser’s analysis flags directly.
• The spam-update overlap. The September spam update ran from September 24 to October 8, far longer than the 2026 updates in March (19.5 hours), June (about 2 days) and August (about 2.7 days). Google describes spam updates as improvements to its automated spam-detection systems and has not said what this one targeted. Some trackers showed anomalies from September 26, so real demotions and bad data can look alike, as Search Engine Journal notes.
• What Google actually penalises. Google’s spam policies define scaled content abuse as mass-producing pages mainly to manipulate rankings, and list using generative AI to create many pages without adding value as one example. The test is value to users, not the tool used.
• A pattern worth watching. Poisoned pages leaned toward Reddit (20.8% to 28.8% of poisoned SERPs), Facebook (28.3% to 33.8%) and LinkedIn (6.2% to 9.3%). Nozzle gives no cause, and we would not read it as a ranking verdict.

Any claim that AI-generated spam caused this is unproven, and the timing does not support it.

What it means for users and businesses

For searchers, there is no evidence of harm so far. If Google is degrading responses only for suspected automation, ordinary users may never see it. That is an assumption to monitor, not a finding.

For businesses, the risk is bad decisions made on bad data:

• False alarms. A tracked page missing from the top 10 on more than half of poisoned pages can read as a ranking collapse in a weekly report.
• Expensive reactions. Rewriting pages, pausing content or replacing an agency over what turns out to be measurement noise.
• Tools that disagree. Vendors often sample rather than count, so some daily swings are noise. One practitioner quoted by Search Engine Roundtable only trusts a movement after spot-checking in an incognito window.
• Uneven exposure. The share of poisoned SERPs varied widely, as the table shows.

Nozzle does not break out B2B software, so check your own keyword history before assuming you sit near the average. Note too that Nozzle’s most recent days may be revised as data fills in.

The impact on SEO, AEO and GEO

SEO. Rank trackers become directional instruments, not ground truth. Search Console remains the better anchor for clicks and impressions, though it has had its own delays, as we noted in our previous article. The practical shift is from “where do we rank today” to “is demand reaching us.”

AEO and GEO. Three points deserve attention:

• AI-visibility tools use the same kind of collection. A vendor’s share-of-voice figure is a sample taken through automated requests, so ask how it is validated.
• SERP features moved inside poisoned data. AI Overviews appeared on 34.9% of poisoned SERPs versus 54.2% at baseline, so a dip may be a data artefact, not a real decline. Separately, DemandSphere reported branded-query AI Overview presence rising from 26.12% on September 1 to 82.06% on September 29. That is one vendor’s measurement taken during the disruption, so spot-check your brand queries before acting on it.
• Off-site presence still carries weight. Buyers and AI answers draw on third-party presence, and the forums and social sites that surfaced more often here are part of that mix. Keep investing there, but do not treat this dataset as proof that it works.

Five questions to ask before you react

1. Did Search Console clicks and impressions for that query change? If not, the drop is probably a data problem.
2. Did it last more than a few days? About 80% of poisoning spells lasted a single collected SERP, and the median recovery was two days.
3. Do other tools agree? Disagreement between vendors is a warning sign.
4. Does the page look thin? Fewer AI Overviews, a missing “People also search for” block, short pagination, unfamiliar sites and extra Reddit or Facebook results are all clues. None proves anything alone.
5. Do business signals agree? Branded search, demo requests and pipeline tell you whether anything real changed.

What B2B marketers should do next

1. Re-anchor reporting. Lead SEO reports with Search Console clicks and impressions plus pipeline, and show rank-tracker data as a directional trend over several weeks.
2. Annotate the window. Mark September 13 to October 9 in your dashboards as a low-confidence period so nobody interprets it later as a performance event.
3. Question your vendors. Ask how they detect failed or poisoned SERPs, what they sample, and how they flag low-confidence days. DemandSphere told PPC Land it is adding error and sampling-rate reporting.
4. Spot-check priority queries. Review your top 20 revenue keywords manually in a clean browser session before escalating any drop.
5. Measure outcomes you control. Pair branded search and direct visits to pricing and demo pages with self-reported attribution on forms, so SEO is judged on pipeline.
6. Re-baseline AI visibility carefully. Use a fixed set of buyer prompts, repeat the runs over time, and report share of appearances instead of a single rank.

What not to do

• Don’t rewrite or disavow after a one-day drop. Wait for persistence and a Search Console signal.
• Don’t tell leadership “Google is poisoning results” as fact. Say vendors report poisoned rank-tracking data and that Google hasn’t confirmed it.
• Don’t blame poisoning for every decline. The spam update was real and finished on October 8. A drop that persists in Search Console after that deserves a proper content and quality review.
• Don’t buy cheap SERP data without asking where it comes from. Google’s litigation against scrapers makes provenance a business risk. The case itself is unresolved.

BuzzIQ’s take

“Google is poisoning search results” is a sticky headline that is only half right. The defensible version is narrower: vendors report that Google answers some automated requests with wrong results, and nobody has shown the same happens to buyers. This is a measurement-trust story, and for B2B teams that is the more useful frame.

Our strategic takeaways:

• Judge SEO on outcomes. Rankings were always a proxy. When the proxy gets noisier, demand, pipeline and win rate carry more weight.
• Build evidence from more than one source. Search Console, CRM, self-reported attribution and manual checks should agree before you change strategy.
• Be precise in the boardroom. Separate what was measured from what was inferred. Teams that do this keep their credibility when the story shifts.

We would also stay humble about the cause. The two reference studies are valuable, but both come from SEO vendors, and Google may yet explain or change its approach. Revisit the question when it does.

Where this applies at BuzzIQ. Our B2B SEO and AI search work includes reporting that holds up when tools disagree. Our AEO and GEO services cover baselining how AI engines describe you. If your dashboard has been alarming you since September, a reporting audit is a sensible first step.

Sources

• Nozzle Research: Poisoned Google SERPs
• Ryan Doser: Poisoned Google SERPs, what they are and how to spot them
• Search Engine Journal (Roger Montti): Analysis, Google is poisoning search results for rank trackers
• Search Engine Roundtable (Barry Schwartz): Third-party Google rank tracking not fully working
• PPC Land: Google returns wrong results on 11% of DemandSphere’s daily rank checks
• Search Engine Journal: Google’s September 2026 spam update is complete
• Google Search Central: Spam policies
• Google: Why we’re taking legal action against SerpApi’s unlawful scraping

Google had not commented publicly on poisoned results as of October 14, 2026. Figures are vendor-measured: Nozzle sells rank tracking, and Ryan Doser discloses that he sells SEO products and uses DataForSEO. Search conditions are changing quickly, so verify numbers before quoting them.