On 22 October 2026, Ahrefs told its customers that Google had been showing suspected search scrapers different results from those seen by real users. The company described the issue as Google “poisoning” search results.
According to Ahrefs, between 5% and 20% of search results are affected, depending on the data provider.
Your website might not have lost any positions at all. Instead, a rank tracker could have collected a version of the search results that real users never saw.
That creates a serious problem for SEO teams. Rankings, dashboards, competitor reports and volatility scores all rely on data collected from search results. If some of that data is wrong, the reports built on it can be misleading too.
Some of the recent reports of extreme Google volatility may therefore reflect problems with the data, rather than changes to actual rankings. The exact split between genuine ranking changes and unreliable data is still unknown.
In this guide, we’ll explain what is happening, how it affects SEO reporting, how to recognise potentially incorrect ranking data, and what content, marketing and development teams should do differently.
We’ll also look at a Python script that can help identify suspicious search results in your own ranking data.
1. What Has Happened?
Ahrefs’ announcement on 22 October confirmed a problem that search data providers had been reporting since mid-September.
Google is not simply blocking requests from suspected scrapers. According to Ahrefs, some requests are receiving complete-looking search results pages that differ from the results shown to real users.
This makes the problem harder to detect. A blocked request or CAPTCHA is an obvious warning. A normal-looking results page filled with different websites can easily be mistaken for a genuine ranking change.
Ahrefs says it is filtering out abnormal results, requesting replacements and cleaning up affected historical data.
One of the most detailed public investigations came from Nozzle, a rank-tracking provider. Its research, published on 10 October and updated on 19 October, examined 27 million results pages collected since 13 September.
Nozzle found that 8.4% of the results pages it assessed were affected, compared with a normal background rate of 0.17%.
Semrush, SE Ranking, AccuRanker and SISTRIX have also acknowledged problems. Their figures vary because each provider collects and assesses data differently.
Google has not made a public statement about the issue.
The widely held view is that Google is targeting AI companies and commercial services that collect its search results, with SEO tools caught up in the process. However, Google has not confirmed this explanation.
What Does a Poisoned Search Results Page Look Like?
According to Nozzle, the problem cannot reliably be identified by looking at a results page on its own. The provider compares each keyword’s results with its own historical data to spot unusual changes.
Several patterns stand out.
The page looks completely normal. There may be no error message, CAPTCHA or obvious layout change. The results page can look just like a genuine Google search.
Most familiar websites disappear. On affected pages, only around one-fifth of the organic websites from the previous day’s results remained. Under normal conditions, almost nine in ten would usually remain.
Social media platforms and forums appear more frequently. Facebook, Instagram, Reddit and YouTube gained the most visibility in the affected results. Government websites and unrelated sites also appeared.
The website being tracked can disappear from the top ten. In Nozzle’s data, a customer’s own website was missing from the top ten on six in ten affected pages. Under normal conditions, this happened on fewer than three in every hundred pages.
That looks exactly like a major ranking drop, even when the website has not actually lost its positions in the results shown to real users.
The changes usually do not last long. Nearly four in five reported incidents affected a single data collection. In most cases, the usual websites returned within three days.
Search features can change as well. AI Overview sources appeared much less frequently on affected pages. Adverts and shopping results, by comparison, were generally more consistent.
This combination of missing websites, unexpected competitors and a rapid return to normal can help identify potentially unreliable ranking data.
However, a sudden change is not proof of data poisoning. A genuine Google update or another major ranking change can produce similar results, so the findings need to be checked against other evidence.
Timeline: How the Problem Developed
| Date | What happened |
|---|---|
| January 2025 | Google began requiring JavaScript for search, disrupting several tools. |
| September 2025 | Google removed the parameter that allowed 100 results per request, making the collection of deeper results ten times more expensive, according to the source analysis. |
| December 2025 | Google sued the search-scraping service SerpApi. The case was still being contested at the time of writing. |
| 13 September 2026 | Requests from search data providers began failing in large numbers. |
| 20 September 2026 | The first affected results pages appeared in Nozzle’s data. |
| 27–29 September 2026 | The problem peaked. On 29 September, 28.6% of Nozzle’s results pages were affected. |
| October 2026 | The reported rate fell to single figures but remained above normal levels. |
| 22 October 2026 | Ahrefs told customers that 5% to 20% of results were affected, depending on the provider. |
The pattern suggests that collecting reliable search results has become increasingly difficult. The reported problem is the latest development in a series of changes that have already made search data more expensive and harder to collect.
Which Searches Are Affected Most?
Nozzle found significant differences depending on the search query, industry and country.
| Category | Reported share of affected results |
|---|---|
| Searches containing “near me” | 12.8% |
| Long-tail searches | 10.0% |
| Brand searches | 5.2% |
| Legal sector | 16.8% |
| Pets sector | 4.3% |
| United Kingdom | 3.6% |
| United States | 8.6% |
| Austria and South Africa | More than 20% |
In Nozzle’s data, “near me” searches were affected most frequently, followed by longer and more specific searches. Brand searches had the lowest reported rate of the three query types.
The Legal sector was the worst affected, while Pets had the lowest rate.
The United Kingdom had the lowest reported rate among the countries listed by Nozzle, at 3.6%. However, that was still substantially higher than the normal background rate of 0.17%.
The type of device used to collect the data made little difference.
These figures come from Nozzle’s analysis and should not be treated as exact rates for every keyword, provider or website.
2. Why Does This Matter to Businesses?
Incorrect ranking data does not necessarily change what real users see in Google. It therefore does not directly cause a website to lose organic traffic.
The problem is that it can make businesses believe something has gone wrong when it has not.
SEO teams use ranking reports to identify problems, measure performance, assess competitors and decide where to invest resources. When those reports contain incorrect information, decisions based on them become less reliable.
1. False Ranking Drops Can Lead to Unnecessary Changes
Imagine your rank tracker reports that an important page has disappeared from the top ten overnight.
Your team investigates, rewrites the page, changes its title and reports the apparent decline to senior management.
But what if the page never lost its real ranking?
If the tracker collected a poisoned results page, the team has spent time fixing a problem that did not exist. Worse, unnecessary changes could create a genuine SEO problem.
One of the most concerning patterns reported by Nozzle is a sudden one-day collapse followed by a recovery within two or three days.
That is a reason to investigate the data before making changes, not a reason to assume every short-lived ranking drop is false.
2. Google Volatility Scores May Contain Incorrect Data
Volatility trackers measure how much search results change from one day to the next.
If Google returns a different set of websites to a data provider, the tracker may record a large change even when the results shown to real users have not changed in the same way.
Nozzle’s reported poisoning peak on 27–29 September coincided with the first wave of the September 2026 spam update and some of the highest UK volatility readings discussed in our analysis of UK search volatility.
Some of that movement may have been genuine. Some may have been caused by unreliable data. At the time of writing, there was not enough evidence to calculate exactly how much came from each.
The problem may also explain why third-party volatility trackers disagreed so sharply in mid-October. If providers collect different results and filter them differently, their measurements of the same period can diverge.
A high volatility score is not, on its own, proof that Google has changed your rankings.
3. Historical Ranking Data May Change
Ahrefs says it is cleaning up affected historical results.
That is a sensible response, but it creates another reporting challenge. A report exported last week may no longer match the same report downloaded today.
If you have used third-party ranking data to analyse the September 2026 spam update, check whether your provider has corrected its historical data.
This is especially important when presenting before-and-after comparisons, explaining a traffic decline or assessing the effect of an SEO project.
Keep previous exports and record when your provider changes its data. That will help you understand why historical reports may no longer match.
4. AI Search Visibility Reports May Understate Results
Nozzle found that AI Overview sources were much less likely to remain on affected results pages.
This means reports built from scraped search results since mid-September may understate how often AI Overviews appear or which sources they cite.
For businesses investing in Generative Engine Optimisation (GEO) and AI search visibility, this matters. A report suggesting that your brand has lost AI Overview citations may reflect a data collection problem rather than a genuine change in Google’s results.
Adverts and shopping results were more consistent in Nozzle’s analysis, so data about those features may be less affected.
Even so, AI visibility reports should be checked against other evidence before they are used to make strategic decisions.
5. Competitor Research Can Become Misleading
SEO teams use ranking data to understand which competitors are gaining visibility, where content gaps exist and how their share of search results is changing.
If a poisoned results page removes established websites and replaces them with social media platforms, the resulting reports can make it appear that competitors have lost ground while Facebook, Reddit or Instagram have gained it.
This can distort:
- Share-of-voice reports.
- Competitor gap analyses.
- Keyword difficulty assessments.
- Competitor traffic estimates.
- Content and keyword prioritisation.
The risk is not simply that a report contains an incorrect number. It is that the business may choose the wrong competitors to target or invest in the wrong opportunities.
6. Collecting Search Data Is Becoming More Expensive
The current problem follows several changes that have made large-scale search data collection more difficult.
Google began requiring JavaScript for search in January 2025. In September 2025, it removed the parameter that had allowed tools to request 100 results at a time. Google also brought legal action against SerpApi in December 2025.
The latest reported development is that some suspected scrapers may receive different results rather than simply being blocked.
These changes increase the challenge of collecting accurate ranking data. As a result, SEO tools may need to adjust how frequently they collect results, how deeply they track keywords and how much they charge customers.
Businesses should watch for changes to their tools’ data coverage, collection frequency and pricing.
The Main Risks at a Glance
| Risk | What it looks like | Potential business impact |
|---|---|---|
| False ranking drops | A page disappears from the top ten for one day and then returns | Unnecessary changes and concern among stakeholders |
| Inflated volatility | A tracker reports extreme market movement | Update analysis based on unreliable data |
| Changed historical data | Old exports no longer match the provider’s latest reports | Reports that cannot be reproduced consistently |
| Underreported AI features | Fewer AI Overview sources appear in the data than users see | Incorrect conclusions about AI search visibility |
| Misleading competitor data | Social platforms suddenly appear to be gaining ground | SEO strategy aimed at the wrong competitors |
The central lesson is that ranking data needs to be checked before it is used to make important decisions.
3. Who Is Affected?
Any business that relies on third-party ranking data could be affected. The level of risk depends on the type of searches being tracked, the provider being used and how heavily the business relies on daily ranking reports.
Local Businesses and Multi-Location Brands
“Near me” searches had the highest reported rate of affected results in Nozzle’s analysis.
That makes local rank tracking and location-based ranking grids particularly important to check. If a tracker suddenly reports that a local service page has disappeared from the top ten, verify the change before adjusting the page or changing the local SEO strategy.
Legal, Finance and Lead Generation Businesses
Legal was the most affected sector in Nozzle’s data.
Businesses in legal services, finance, insurance and lead generation often depend on detailed ranking reports to understand their position in competitive markets. Agencies and in-house teams reporting weekly results to clients, partners or senior management should be particularly careful about false drops.
E-commerce Businesses
Long-tail product and category searches were affected more frequently than brand searches.
Shopping and advert data appeared to hold up better, which suggests that the biggest concern for retailers is organic ranking data rather than every type of search feature.
Retail teams should still check sudden ranking changes against Search Console and other reliable evidence before changing category pages, product content or SEO priorities.
Publishers and AI Search Teams
AI Overview source data appears to be among the more heavily distorted areas.
Teams reporting how often Google AI Overviews cite their websites should compare third-party reports with first-party data and manual checks.
A fall in reported citations should not automatically be treated as a loss of AI visibility.
Businesses Collecting Their Own Search Results
Companies using their own APIs or web scrapers are not automatically protected from this problem.
In fact, smaller teams may lack the filters and validation systems that larger SEO tools are now developing.
If your business collects search results directly, you should check for unusual changes in the results pages, retain the original responses and validate important ranking movements against your own site performance.
UK Businesses
The United Kingdom had the lowest reported rate of affected results among the countries in Nozzle’s comparison.
However, its reported rate of 3.6% was still around 21 times the normal background rate of 0.17%.
UK businesses should therefore treat this as a relevant data-quality issue, even if the reported impact is lower than in some other markets.
4. What Should Businesses Do Now?
You do not need to stop using rank trackers. You need to stop treating every daily ranking result as an unquestionable fact.
Use Google Search Console to understand how your own website is performing, validate unusual changes before acting on them, and update the rules your team follows when a ranking drop is reported.
The following steps explain how to put that into practice.
Step 1. Make Search Console Your Main Source of Truth
Who: SEO lead
Google Search Console reports your website’s performance in Google Search. A poisoned results page collected by a third-party tracker does not directly change the impressions and clicks Google reports for your site.
When a rank tracker and Search Console appear to disagree, investigate the Search Console data before accepting the ranking change as real.
For example, if your tracker says a page has disappeared from the top ten but Search Console still records impressions for the relevant query on that day, this is a reason to investigate whether the tracker collected an abnormal results page.
Use daily query-level data where available, rather than relying only on 28-day totals. Search Console’s bulk data export to BigQuery can help you retain daily query and page performance data. Our bulk data export guide explains how to set it up.
However, Search Console has limitations of its own:
- Average position is calculated using impressions, so it does not represent a simple daily ranking for a keyword.
- Some rare queries are anonymised.
- The newest data can be provisional.
Search Console is an important source of evidence, but you still need to understand what its figures do and do not show.
Step 2. Change How Ranking Drops Are Investigated
Who: SEO lead and marketing
Do not make major decisions based on a single ranking collection.
Use the following process when a tracker reports a sudden decline.
1. Wait for confirmation. A ranking drop should normally appear on at least two consecutive days before it is treated as an ongoing issue. Do not delay fixing an obvious technical fault, but avoid changing content based on one unexplained ranking report.
2. Look for unusual results. Check whether many of the keyword’s familiar websites have disappeared at the same time. Look for unexpected increases in social media platforms and forums, unusually short results pages, and a rapid return to the previous results.
3. Check Search Console. Before escalating the issue or changing the affected page, establish whether your own site’s clicks and impressions support the claim that something has changed.
4. Allow time for the data to recover. If there is a sudden decline that Search Console does not support, monitor it for up to three days. Most of the reported poisoning incidents cleared within that time.
These checks are not a guarantee that every short-lived decline is false. Genuine ranking changes can also happen quickly. The purpose is to gather enough evidence before taking action.
Step 3. Ask Your SEO Tool Provider the Right Questions
Who: Marketing and procurement
Your SEO tools should explain how they detect and handle abnormal search results. Ask your provider the following questions.
| Question | Why it matters |
|---|---|
| What proportion of your search collections are being rejected or replaced? | Helps you understand how much of your collected data may have been affected. |
| How do you identify abnormal results pages? | Comparing results with each keyword’s own history is one method supported by the public research. |
| Have you corrected historical data? If so, from what date? | Helps you identify which old reports and exports may need to be checked again. |
| Are volatility scores calculated before or after abnormal results are filtered out? | Shows whether the market volatility report may include affected pages. |
| Do you tell customers when data collection is disrupted? | Helps you understand whether reporting issues will be flagged clearly. Ahrefs says it plans to use dashboard notifications. |
| How frequently do you collect rankings, and how deep do you track results? | Collection frequency and ranking depth may change as the cost of gathering data rises. |
The aim is to understand what your provider collects, what it filters out and how it communicates problems.
A supplier that explains its methods and reports data-quality issues clearly gives you more information to assess the reliability of its reports.
Step 4. Recheck Recent SEO Analysis
Who: SEO analysts
Review recent reports that rely on third-party ranking data collected from 13 September onwards.
Start by exporting the latest data after your provider has completed any clean-up. Compare it with the figures you previously reported and note any differences.
Then review the following areas:
Google update analysis. Treat analysis of the September 2026 spam update and recent volatility as provisional where it relies on third-party rankings that may have been affected.
Competitor visibility. Re-run share-of-voice reports and check any sudden gains by Facebook, Reddit, Instagram or YouTube. These could reflect genuine changes, but unexpected gains may also be a sign of abnormal results.
AI Overview citations. Compare third-party reports with other available evidence, including Search Console’s AI performance data, manual checks in a normal browser and referral traffic in your analytics platform.
Reports shared with leadership or clients. If historical numbers have changed, explain why. Keeping the old and updated figures together makes it easier to understand what changed and whether earlier decisions need to be revisited.
Step 5. Filter Your Own Search Data
Who: Developers
If your team collects search results directly through an API or its own scraper, consider adding a filter to identify suspicious results pages.
The poisoned_serp_filter.py tool described below uses a keyword’s historical results to flag snapshots where most of its usual websites have disappeared. It can write a cleaned file without the flagged snapshots and compare suspicious results against your own Search Console data.
There are four additional practices worth following.
Record the details of every collection. Keep the date and time, location, device, number of results pages collected and response status. An unusually short or incomplete results page may be a warning sign.
Keep the original responses. Do not immediately overwrite the raw data with a cleaned version. Retaining the original responses allows you to review them again if your detection methods improve.
Collect a fresh result when something looks wrong. Re-request flagged keywords rather than automatically using the suspicious result in your reporting.
Follow Google’s and your provider’s terms. Trying to get around detection systems can create legal and commercial risks. It is better to focus on validating your data than attempting to evade restrictions.
Step 6. Add a Reporting Note and Explain the Situation
Who: Everyone
Add the following annotation to your reporting:
13 September 2026 onwards: third-party search results data may be affected by abnormal results.
Keep this alongside annotations for confirmed Google updates.
Tell leadership and clients about the issue before the next unexplained ranking drop appears in a report. Explain that some third-party ranking data may be unreliable, how your team will check unusual movements, and which sources will be used to confirm genuine changes.
This is better than allowing stakeholders to draw conclusions from an unexplained ranking report.
Step 7. Give Each Team Clear Responsibilities
Every team should understand how the issue affects its work and what to do when suspicious data appears.
| Team | What changes | First actions |
|---|---|---|
| Content | Page changes require evidence from Search Console, not just a tracker alert. | Pause rewrites prompted by single-day ranking drops since mid-September until the changes have been checked. |
| Marketing | Reports need a data-quality note, and suppliers need to explain their filtering methods. | Ask providers the questions in Step 3, annotate reports and brief stakeholders. |
| Development | Data collection needs its own checks. | Add a filter to in-house collection systems, retain raw responses and build a Search Console cross-check. |
| Leadership | Daily ranking data may be less reliable for a period. | Agree escalation rules based on repeated changes, Search Console evidence and a monitoring period. |
The objective is not to abandon ranking data. It is to make sure that everyone knows when it can be trusted and what evidence is needed before acting on it.
5. Developer Toolkit: Filtering Suspicious Google Results
The poisoned_serp_filter.py script applies a published method for identifying potentially poisoned search results to your own ranking data.
It uses Python’s standard library and reads the same CSV format as the volatility script in our UK search volatility analysis:
datekeywordpositionurl
The script compares each keyword’s results with a baseline period that you choose from before 13 September 2026.
How the Filter Works
The process is based on Nozzle’s published method, adapted for use with your own ranking data.
1. Establish which websites normally appear.
Choose a baseline period before 13 September 2026. A website is considered a usual result for a keyword if it appeared in the top ten on at least 60% of the baseline snapshots.
Each keyword needs at least five baseline snapshots. Keywords with fewer observations are excluded from this calculation.
2. Check each later results page.
For each subsequent snapshot, the script checks how many of the keyword’s usual websites still appear anywhere on the collected page.
It flags the snapshot when fewer than half of the usual websites are present.
3. Report and filter the results.
The script reports the proportion of flagged snapshots for each day. If you use the --out option, it writes a cleaned file containing the rows from unflagged snapshots, ready for further analysis.
4. Compare the result with Search Console.
With the --gsc option, you can supply a CSV containing dates, queries and impressions from Search Console. With --domain, the script can check how often your site disappeared from a flagged results page while Search Console still recorded impressions.
This provides an additional reason to suspect that the ranking snapshot is unreliable rather than evidence of a genuine ranking loss.
Example Command
The following command shows how to run the script using a ranking export, a baseline date range, a Search Console export and your website’s domain.
python3 poisoned_serp_filter.py rankings.csv \
--baseline 2026-08-20:2026-09-09 \
--out rankings_clean.csv \
--gsc gsc_daily.csv \
--domain example.co.ukReplace the example filenames, dates and domain with your own data.
The Main Python Functions
The following code shows the main part of the filtering process. It is an excerpt, not a complete standalone script. The full programme also requires the data-reading functions, configuration values and supporting code.
def usual_sites(snap, start, end):
per_kw = defaultdict(list)
for (kw, date), s in snap.items():
if start <= date <= end:
per_kw[kw].append(s["top10"])
usual = {}
for kw, tops in per_kw.items():
if len(tops) < MIN_BASELINE:
continue
counts = defaultdict(int)
for t in tops:
for s in t:
counts[s] += 1
sites = {
s for s, c in counts.items()
if c / len(tops) >= USUAL_SHARE
}
if sites:
usual[kw] = sites
return usual
def judge(snap, usual, end):
"""Return the kept share and flag for snapshots after the baseline."""
out = {}
for (kw, date), s in snap.items():
if date <= end or kw not in usual:
continue
kept = len(usual[kw] & s["all"]) / len(usual[kw])
out[(kw, date)] = (kept, kept < KEEP_SHARE)
return outImportant Limitations of the Filter
The filter can help identify suspicious results, but it cannot prove that Google has returned incorrect data.
A genuine, large ranking change can also remove many familiar websites from a results page. A core update, for example, could cause a substantial change in the results for a keyword and trigger the filter.
For that reason, check flagged dates against confirmed Google updates and your own Search Console data before deciding that the results are false.
The baseline also matters. If it is too old, it may no longer represent the websites that normally rank for a keyword. Review and refresh it when the reported problem has settled.
The script should therefore be used as a screening tool, not as an automatic way to decide which ranking data is correct.
6. What We’re Watching Next
Several developments will help clarify how serious the issue is and what it means for SEO reporting.
Will Google Make a Public Statement?
Google has not made a public statement about the reports.
An explanation of whether Google is deliberately returning different results to suspected scrapers, and why, would help clarify what is happening and which services are affected.
Until then, the intended target remains unconfirmed.
Will the Problem Settle or Get Worse?
The reported rate of affected results fell from its late-September peak to single figures during October, but it had not returned to the previous background level.
If the rate increases again, it could make future Google update analysis even harder to interpret.
SEO teams should continue comparing third-party ranking data with their own website performance rather than assuming the problem has disappeared.
How Will SEO Tools Respond?
SEO providers may introduce more filtering, reduce collection frequency, collect fewer results per keyword or increase prices as the cost of collecting data rises.
Providers that publish their rejection rates and explain how they identify abnormal results will give customers more information about the reliability of their data.
It will also be important to understand whether filtering happens before volatility scores are calculated and whether historical results have been corrected.
Will Historical Volatility Scores Be Revised?
As providers clean up past data, some of the volatility readings from late September may be revised downwards.
We will compare the updated figures with the original readings in our volatility tracker comparison.
This will help distinguish changes in the actual search results from changes caused by unreliable data collection.
What Happens With the SerpApi Case?
Google’s legal action against SerpApi may help clarify the legal issues around collecting search results.
The case was still being contested at the time of writing, so its eventual outcome remains uncertain.
Will First-Party Data Become More Important?
As third-party ranking data becomes harder to collect and validate, data from Google Search Console, analytics platforms and your own logs becomes increasingly valuable.
We also expect continued pressure for Google to provide more useful AI search performance data. At the time of writing, the AI performance reporting discussed in this article showed impressions but did not provide clicks and queries in the same way as traditional Search Console reporting.
For SEO teams, the practical response is clear: use third-party tools to understand the market, but rely on your own website data to confirm what is happening to your business.
About Szymaniak Digital
Szymaniak Digital is a UK enterprise SEO and AI search consultancy. We help businesses measure organic search performance accurately, check third-party ranking and volatility data before it reaches management reports, and distinguish genuine ranking losses from problems with the data.
If your rank tracker has shown unexplained drops since mid-September, we can help compare them with Search Console, investigate the cause and improve your reporting process so that suspicious changes are checked before anyone acts on them.
Get in touch with Szymaniak Digital to discuss your SEO reporting and AI search measurement.
