SpamBrain
Google's AI-based spam-detection system. It demotes scaled, near-duplicate, low-value content — often site-wide, with no manual action.
SpamBrain is Google's machine-learning system for detecting spam and low-value content. It runs continuously and can demote pages — or an entire site — without any human reviewer and without a manual-action notice ever appearing in Search Console. The traffic simply falls, which makes the cause hard to diagnose after the fact.
The pattern SpamBrain is especially good at catching is scaled content abuse: large numbers of pages that are near-duplicates of one another, where the only real variation is a swapped keyword, a query-string parameter, or a templated block repeated across siblings. A site can generate hundreds of such URLs in a single deploy and look fine for days before the classifier reclassifies the whole cluster as thin, auto-generated, or unhelpful.
This exact failure — a batch of query-string near-duplicate URLs quietly demoted site-wide — is the reason this product runs every content and technical change through a safety linter first. The rules that block near-duplicate bodies, query-string sitemap URLs and over-fast publishing are all encodings of "don't give SpamBrain a reason."
Related terms
- Canonical Tag — A rel="canonical" link that names the preferred URL for a page, telling search engines which version to index when duplicates exist.
- Core Update — A broad, periodic change to Google's ranking systems. Rankings shift site-wide; recovery comes from improving overall quality.
- Keyword Cannibalization — When several pages on one site target the same query and compete with each other, splitting signals and confusing engines.
- Thin Content — Pages with little unique value — auto-generated, scraped, doorway or near-duplicate pages that spam systems demote.