Poisoned SERPs, why your rank tracker shows drops that never happened
Over the past few weeks, our own site has shown sudden ranking drops in Ahrefs that made no sense. Pages that had held their positions for months vanished from the top 100 for a day and then came back. When I log in, Ahrefs now shows a notice that its ranking data is volatile and that it is working on a fix.
We are not alone, and the reason now has a name.
AccuRanker calls them poisoned SERPs. If you report rankings to a board, a client or yourself, you need to understand them before you react to the next drop.
Ahrefs volatility
What is a poisoned SERP?
A poisoned SERP is a Google results page that looks normal and shows the wrong organic results. The ads and AI Overviews on the page stay the same. Only the organic listings change, sometimes to a different set of sites, sometimes to a cut-down page with few results and, for local searches, sometimes to businesses from the wrong city.
A person searching on Google rarely sees one. Rank trackers, which check thousands of keywords automatically, see them far more often. The result is a ranking report that shows a crash that never happened.
AccuRanker's example sums it up. A site ranks third on Monday, Tuesday and Thursday. On Wednesday it disappears from the top 100, and only three of its usual ten competitors appear alongside it. The next day everything is back to normal.
How rank trackers collect their data
To see why this happens, it helps to know how rank trackers work.
Tools such as Ahrefs, AccuRanker and Semrush run your keywords through Google automatically, often thousands of times a day, from different locations and devices. They read the results page each time and record where your site appears. That gives you a daily position for every keyword, which Google itself does not offer.
Google Search Console works differently.
It reports what real people saw when they searched, averaged across every impression your pages earned. It runs a couple of days behind, and its average position blends many searches into one number. It will never match a rank tracker exactly, and it was never meant to. When the two disagree sharply on the same day, though, Search Console is the one that reflects real searchers.
The two sources do different jobs.
A rank tracker tells you where you stand on keywords you choose, every day. Search Console tells you how Google actually showed your site to real people. You need both, and right now you need to trust the second one more.
How big is the problem?
AccuRanker's data, published on 19th October 2026, shows how quickly it grew.
In 2025, whole-page flips were rare, at around one in 1,000 daily checks or fewer.
From 10th to 16th September 2026, the rate sat at about 0.26% on desktop and 0.38% on mobile.
On 25th September it jumped to 30.2% on desktop and 25.7% on mobile.
From 1st to 9th October it averaged about 1.47%, and by 17th October it had fallen to around 0.5%.
Another rank tracker, Nozzle, measured a peak of 27% on 29th September. More than nine in ten poisoned results lasted a single day, and a second check from a different browser session returned normal results about three times in four.
Share of rank checks that returned a poisoned SERP
Percentage of AccuRanker's daily checks, desktop and mobile
Source: AccuRanker, 19th October 2026. The 1 to 9 October figure is the average across both devices.
Why is Google doing this?
Google has not explained it. AccuRanker suspects it is part of Google's effort to stop bulk scraping of its results by AI answer engines, with rank trackers caught in the crossfire.
The poisoning appears to follow the browser session collecting the data, not the keyword, which fits that theory.
The timing makes things harder to read. Google also confirmed its September 2026 spam update, which began rolling out on 24th September, the day before poisoning peaked. So some sites saw real ranking changes in the same week their trackers were serving bad data. Telling the two apart matters.
| Real ranking drop | Poisoned SERP | |
|---|---|---|
| How long it lasts | Days or weeks, often sliding gradually | Usually one day, then back to normal |
| Competitors | Your usual competitors still appear | Many usual competitors vanish at once |
| Search Console | Average position and clicks fall too | Average position and clicks barely move |
| Local results | Local rivals move around you | Businesses from other cities appear |
| A second check | Shows the same drop | Usually shows normal results |
What our own Search Console showed
When Ahrefs showed our pages dropping, I checked Google Search Console, which reports what real searchers saw.
On 25th September, the day poisoning peaked, our site recorded 9,095 impressions and an average position of 10.3, our best figures of the month. Across late September and early October, our average position stayed within its usual range of roughly 12 to 17.
blumint.co average position in Google Search Console
10th September to 17th October 2026, lower is better. The pink point marks 25th September, when poisoning peaked.
Source: Blu Mint's Google Search Console data.
The drops in the tracker were not happening in Google's own data. That is exactly the pattern AccuRanker describes, and it is why I never act on a single day of rank tracking alone.
The real cost of reacting to bad data
A fake drop costs nothing on its own. The damage comes from what people do next.
A marketing manager sees an important page fall out of the top 100 and rewrites it. A founder sees a sudden slump and questions whether the agency is doing its job. A team changes title tags, removes content or moves budget out of SEO, all because of one day of poisoned data.
Changing a page that was never in trouble can create the very drop you feared.
Google has to recrawl and re-evaluate the page, and the new version may not perform as well as the old one did. Panic moves also make the data harder to read later, because you no longer know whether a ranking change came from Google, the poisoning or your own edits.
The better habit is to wait, check and only then act. One bad day is a question, not an answer.
Will AI search tracking face the same problem?
Very likely, and in some ways it already does.
AI citation tools work like rank trackers: they run your target questions through ChatGPT, Gemini, Perplexity and Google's AI features and record whether your brand appears. AI answers already vary from one run to the next, so a single check tells you little.
In my experience, standalone AI citation software is not always accurate.
I get a more useful picture from Ahrefs' Brand Radar alongside paid subscriptions to the most capable AI models, checking the same questions over time and looking at the trend. If AI platforms start fighting automated checks the way Google appears to be, citation data will need the same caution as rankings do now.
What to do before you panic
If your rank tracker shows a sudden drop, work through these checks first.
Compare the same query and date in Search Console. If position and clicks have not changed, treat the tracker's drop as bad data.
Wait a day. Most poisoned results correct themselves within 24 hours.
Look at your usual competitors. If most of them disappeared too, you are looking at a poisoned page.
Judge rankings on trends over a week or more, not on one day.
Ask your rank tracker how it detects poisoned pages and whether it re-checks them before reporting.
This is also another reason to stop treating rankings as the main measure of success. AI Overviews already take clicks from the top result, and now the ranking data itself is less reliable. We report on Search Console impressions, AI citations, branded searches and enquiries alongside rankings, as I explain in our article on AI citation tracking. Our guide to choosing a good SEO expert covers why honest reporting matters, and ChatGPT's own search index shows how far search now reaches beyond Google.
If a sudden drop has you worried, our SEO team can tell you whether it is real. And if organic traffic needs support while you find out, paid search can fill the gap, backed by content, a well-built website and a steady social media presence.
How we report rankings now
This is also another reason to stop treating rankings as the main measure of success.
AI Overviews already take clicks from the top result, and ranking data is now less reliable.
We report on Search Console impressions, AI citations, branded searches and enquiries alongside rankings, as I explain in our article on AI citation tracking.
Each month our client reports cover the following.
Rankings from Ahrefs, read as trends over weeks and checked against Search Console.
Search Console impressions and clicks, the closest thing to what real searchers saw.
AI citations and brand mentions, tracked with Brand Radar and our own checks.
Branded searches and direct visits, which show growing awareness.
Enquiries and revenue, which are what the work is ultimately for.
No single number tells the whole story, and that has never been more true than it is this autumn. Our guide to choosing a good SEO expert covers why honest reporting matters, and ChatGPT's own search index shows how far search now reaches beyond Google.
If a sudden drop has you worried, our SEO team can tell you whether it is real. And if organic traffic needs support while you find out, paid search can fill the gap, backed by content, a well-built website and a steady social media presence.
Ranking data itself has become less reliable this autumn. Google has started serving fake results pages to automated checks, so rank trackers can report drops that never happened. That's another reason to track visibility, citations and enquiries alongside rankings.
-
A poisoned SERP is a Google results page that looks normal and shows the wrong organic results, while the ads and AI Overviews stay the same. Rank trackers see them far more often than real searchers, which creates ranking drops that never happened.
-
No, they affect the data, not your rankings. Real searchers still see your normal position. Check Google Search Console for the same query and date to confirm.
-
AccuRanker and Nozzle have both published data on the problem, and Ahrefs has flagged volatile ranking data. Google can serve poisoned pages to any tool that checks its results automatically, so ask your provider how it handles them.
-
Google has not said. The leading theory is that it is trying to stop AI companies scraping its results in bulk, and rank trackers are caught up in it.
