GSC Data Indicates Scaled Content Classifiers Kill 99% of Impressions Without Deindexing URLs
A medical site expanded from 280 to 1,872 pages in July via an AI-translation pipeline, triggering a catastrophic ranking collapse.
Impressions flatlined from 48,000 to 652, and, during the Google's August 2026 spam update, the average position sank from 24.9 to 52.8 overnight.
The algorithm executes this Scaled Content Abuse penalty at the query level, bypassing standard deindexation.
Forensic checks confirm all 40 heavily suppressed pages pass URL Inspection and remain explicitly indexed in GSC.
Recrawling the URLs fails to reverse the suppression.
While the site retained 1,450 indexed URLs, active queries plummeted from 32,898 to 6,966, indicating the classifier blocks query visibility rather than removing the assets from the database.
Analysts trace the algorithmic strike to two concurrent triggers intersecting with the August spam update.
First, the programmatic expansion injected 230 translated pages in each of the eight languages simultaneously.
Second, the domain fabricated YMYL authority signals by generating fake physician profiles using ChatGPT o4 and Google Imagen 4 to simulate medical review.
Practitioners confirm that algorithmic classifiers target the generation methodology itself; recovery requires first halting deployments and reverting or noindexing the worst-performing cohort, rather than touching backlinks, author entities or sitewide templates before a full recrawl cycle.