Ahrefs says it is scrubbing historical rank data it believes Google distorted, and four other vendors have now published how they screen suspect results. Search Engine Journal’s Matt G. Southern compared those statements in a report dated Friday, October 23, 2026. Every claim comes from a vendor describing its own system.
Earlier this month, we covered Nozzle’s claim that Google hands suspected scrapers swapped-out pages, and Morningscore’s report of truncated results. This is the next turn: what the tool makers say they are doing about it.
Ahrefs posted its update on Thursday, October 22. It says roughly 5% to 20% of results were hit in recent weeks, with the share varying by data supplier, and that its filters discard abnormal pages and ask for fresh ones. Ahrefs also says it is “cleaning up any past results that may have been affected.” Southern notes the post names no dates, reports, or metrics for that cleanup. Ahrefs uses the word “poisoning” for Google’s behavior. That is the vendor’s characterization, and Google has not commented.
SEOmonitor gives the most specific timeline. Its help center says some wrong results were stored between October 9 and 13, and that it reprocessed the affected keywords on October 13, when a new check went live. It also says it reprocessed keywords hit by bad data from September 25 to 28. By its account, about 10% of monitored keywords had problems in the October window, with most Visibility movements small. It adds that around 8% of daily results are incomplete or wrong and get pulled again before storage.
AccuRanker describes a history comparison. It discards shortened pages, fetches again in a separate browser session, and checks page-one sites against recent days. If fewer than half the usual sites appear, it retries, and a repeat of the same new sites counts as a genuine change. About one in 20 desktop pages and one in 12 mobile pages lack enough history for this check. Its API product, AccuSERP, cannot run it at all, so it favors browser sessions that passed earlier screening.
BrightLocal takes a supply-side approach. CEO Myles Anderson wrote on October 16 that the company is working with its current data suppliers and adding more, so one failing source does not stall the product. He also said “no single fix that solves it completely” exists. SerpApi, a vendor that sells Google results by API, said on October 14 that a deployed update “should reduce” bad results and that work continued.
Nozzle published a companion piece on October 22, “Google Plays Dumb,” with data through October 21. It describes a separate fault: Google answers only one word of a multiword query, so listings drift off topic while the page title keeps the whole phrase. Nozzle reports this hit 13.9% of judged SERPs from September 16 on, peaking at 39.9% on September 28. It says a tracked site dropped from the first ten results on 69.8% of those pages, yet the next check within three days looked normal in 85.8% of cases.
Here is the framing the vendor posts lack: the numbers cannot be compared across companies. SEOmonitor’s 8% counts results Google returned. AccuRanker’s flag rates come from a sample of one in 200 keywords. Ahrefs does not say whether its range is measured before or after filtering. Southern flags the same mismatch and notes that no vendor points to independent testing of its checks.
Handling of unconfirmed data differs too. SEOmonitor says that for large drops it may keep showing the earlier position for as many as two daily updates, until the lower position repeats. AccuRanker does not say what its dashboard shows during a re-fetch. Clients of different tools could see different charts for the same October week.
Agencies should ask each vendor in writing whether stored values for late September and October were revised, and on which dates. Exports sent before a cleanup may no longer match the live tool, so compare old and new numbers before presenting that period to a client, and keep Search Console as the second source for any drop.
Search Engine Journal, Matt G. Southern, October 23, 2026, drawing on Ahrefs’ October 22 post and Nozzle’s “Google Plays Dumb” research.
