Tracking Google Anomalies

SEO tracking providers are explaining how they identify unusual Google search results, check their accuracy and decide what data customers see. Several companies have also outlined their plans for reviewing previously recorded rankings that may have been affected by irregular results.

Ahrefs recently announced that it was reviewing historical data potentially affected by what it describes as search result “poisoning”. However, the company has not specified which dates, reports or metrics will be included in the review.

SEOmonitor has also acknowledged that some inaccurate results were recorded between 9 and 13 October. The company says it reprocessed affected keywords on 13 October, when it introduced an additional validation check.

Other providers, including AccuRanker, BrightLocal and SerpApi, have shared details of their own responses. Their approaches differ in how they verify results before saving them, display rankings while checks are ongoing and deal with potentially inaccurate historical data.

What Nozzle found in its research

Rank-tracking company Nozzle published research on 10 October examining Google results pages that appeared to be missing many of the websites normally shown for particular search terms. The report was updated on 21 October.

Nozzle classifies a first-page result as potentially affected when fewer than half of the websites normally appearing on that page are found anywhere in the collected results. Its method does not require a subsequent check to confirm that the missing websites have returned.

Using this measure, Nozzle estimated that 8.7% of the first-page results it analysed were affected between 13 September and 21 October. For comparison, the figure was 0.17% between 20 August and 9 September.

The company believes the pattern is connected to Google’s efforts to limit AI-related scraping, leaving rank-tracking providers affected by the changes.

How providers verify results before recording them

Rank trackers use different checks to determine whether a sudden ranking change reflects a genuine movement or an unreliable Google response.

AccuRanker says it first filters out shortened or incomplete search results pages before requesting the results again through a separate browser session. It then compares the new first page with the websites that appeared on the first page for the same keyword on at least two of the previous three days.

If fewer than half of those websites appear, the system requests the results again. When the second check returns the same new set of websites, AccuRanker treats the change as genuine.

However, the company notes that its historical comparison cannot be applied to every search. Around one in 20 desktop results pages and one in 12 mobile results pages lack sufficient historical data for this check.

SEOmonitor estimates that approximately 8% of the Google results it receives each day are incomplete or inaccurate and need to be retrieved again before being recorded. The company has not specified the period covered by this estimate.

On 13 October, it introduced a new process that compares incoming results with the recent history of the relevant keyword.

Ahrefs also says it uses filters to identify unusual results, reject them and request replacements. It estimates that between 5% and 20% of search results may have been affected in recent weeks, depending on the data provider. Its announcement does not clarify whether this estimate refers to results before or after its filtering process.

BrightLocal is taking a different approach by working with existing data suppliers and adding further providers. This is intended to offer alternatives if one source experiences problems. Its chief executive, Myles Anderson, has stressed that there is no single solution that completely resolves the issue across the industry.

What customers see while rankings are checked

The rankings displayed in an SEO platform depend not only on the data received from Google but also on the provider’s verification procedures and display rules.

SEOmonitor has explained that it may retain a previously recorded ranking for up to two daily updates when it detects an unusually sharp decline on the same device. For example, this could involve a keyword dropping from fourth to 14th place or disappearing beyond the top 50 results.

The previous ranking remains visible while the platform checks whether the decline persists. SEOmonitor also provides a recrawling option and a Chrome extension that allows users to compare rankings.

AccuRanker says that when its historical comparison identifies a page with an unusually large change, it retrieves the results again before sending the data to the customer’s dashboard. The company has not explained precisely what users see while this second check is taking place.

Users can also request a keyword refresh directly through the platform, prompting a new search through a separate browser session.

What happens to rankings already recorded?

Correcting new results does not necessarily resolve inaccuracies that have already entered a platform’s historical records. Providers have therefore outlined different approaches to reviewing earlier data.

SEOmonitor says it reprocessed keywords affected by inaccurate results between 25 and 28 September. It also reports that it prevented incorrect data from entering its platform until 8 October.

However, the company says that from around 9 October, many questionable results no longer displayed the signals its original quality checks relied upon. Between 9 and 13 October, approximately 10% of its monitored keywords were affected by inaccurate results, although most movements in its Visibility metric were reportedly small.

Its announcement does not specify which historical daily rankings were subsequently amended.

AccuRanker uses what it calls a “sandwich check” to identify potential anomalies in stored data. The process looks for days when fewer than half of the websites appearing on the first page match those recorded on the preceding and following days.

The company applied this check to a fixed sample of one in every 200 keywords. It reported that 30.2% of sampled desktop keyword-days were flagged on 25 September, compared with 0.58% on 17 October.

However, AccuRanker has not confirmed whether flagged historical entries will be amended in customers’ stored data.

How Google results API providers are responding

The issue also affects businesses that obtain search results through APIs rather than using a conventional rank-tracking dashboard.

On 14 October, SerpApi reported that it had introduced an update intended to reduce the number of potentially affected Google results returned by its service. The company said it was continuing to work on the problem.

A customer had also requested a field that would identify searches returning irrelevant results. A comment in the same public discussion noted that services using the API currently have to assess each response themselves.

AccuRanker’s API product, AccuSERP, returns the results Google provides for an individual search and cannot perform the same historical comparison as the company’s rank tracker.

AccuRanker says the API shares a pool of browser sessions with its main tracking service and prioritises sessions that have passed its validation checks and continued to return reliable results.

Why SEO teams should review their data

Agencies investigating ranking drops recorded in September and October should first establish how their tracking provider handled results during the affected period.

For example, SEOmonitor may temporarily retain an earlier ranking while checking whether a sharp decline is genuine. AccuRanker users can request a fresh keyword check, while other validation methods can help establish whether the websites normally appearing on the first page have disappeared.

Google Search Console can provide additional context when investigating ranking changes. Its performance reports calculate average position using the highest-ranking result from a website for a given search, and Google notes that results can vary according to device, location, timing and search history.

A difference between Search Console data and a rank-tracking platform does not automatically prove that either source is incorrect. The figures measure results in different ways and may reflect different search conditions.

Before including September or October figures in client reports, agencies should ask their providers whether historical records have been corrected. If data has been updated, older exports may no longer match the figures currently shown in the platform.

Businesses that retrieve Google results through APIs should also review the safeguards available at their chosen endpoint, including session screening and historical comparisons. Internal systems may need additional validation before accepting responses as accurate.

Questions remain unanswered

As of 22 October, none of the providers covered in the reports had cited a public statement from Google explaining the changes. The companies have described their own systems, but none has provided independent testing demonstrating how effectively its checks identify inaccurate results.

The percentages published by different providers also measure different things and should not be treated as directly comparable. SEOmonitor’s estimate concerns the results it receives from Google, while AccuRanker’s figures relate to a sample of stored keyword-days.

SEOmonitor says it will continue updating its incident page as Google’s changes develop and its own adjustments are introduced. Ahrefs has also said it will publish updates through its dashboard.

For SEO teams, the key step is to verify how their provider handled potentially affected data before relying on historical rankings in client reports or management decisions.