IP-Prior: How Google Protects Click Data from Manipulation
The system that detects unnatural click patterns and protects NavBoost from gaming — the integrity layer of behavioral ranking
01What is IP-Prior?
IP-Prior protects the integrity of click data. It detects unnatural click patterns, for example many clicks from the same IP addresses or clicks artificially directed at a page. Squashing is the mechanism that reduces the weight of suspicious clicks. The voterToken system adds a layer that verifies legitimate clicks.
The Google API leak confirmed an IP-Prior field with the revealing name unscaledIpPriorBadFraction. It holds the fraction of clicks from an IP address that are classified as 'bad' before squashing. A high bad fraction means many clicks from that IP look manipulative, and their weight in NavBoost is reduced.
IP-Prior feeds into NavBoost (System 1), the system that evaluates clicks from search. It is the integrity layer that filters NavBoost's input. Without IP-Prior, NavBoost would be vulnerable to click farms and automated clicking. With IP-Prior, manipulative clicks are detected and downweighted before they can influence rankings.
02How squashing works
Squashing is the mathematical mechanism that reduces the weight of suspicious clicks. It is not a binary filter: suspicious clicks are not removed, they are downweighted. A click that is 80% likely to be manipulative might have its weight reduced to 20% of its original value. This gradual approach keeps false positives from removing legitimate clicks completely.
The voterToken system adds another layer of verification. A voterToken is a cryptographic identifier that verifies a click came from a legitimate user session. Clicks without valid voterTokens, or with tokens that look machine-generated, receive lower weights in the squashing process.
IP-Prior, squashing, and voterToken together form a three-layer defense. IP-Prior detects suspicious patterns, squashing reduces their weight, and voterToken verifies legitimacy. This layered approach is characteristic of Google's anti-spam architecture: each layer catches what the previous one missed.
03Implications for SEO practitioners
IP-Prior means that click manipulation is detectable and actively filtered. Click farms, automated clicking tools, and coordinated click campaigns will be detected. They are not necessarily penalized, but squashing neutralizes their effect on NavBoost.
Legitimate SEO tactics that generate real clicks are therefore safe: improved titles, better snippets, featured snippet optimization. IP-Prior does not filter genuine user behavior. The system is designed to distinguish real engagement from artificial engagement, not to penalize all engagement.
Because of the three-layer defense, manipulating one layer is not enough. A click farm could use rotating IP addresses (defeating IP-Prior), but the clicks would still lack valid voterTokens (caught by the verification layer). Or it could generate valid tokens, but squashing would detect the click patterns as unnatural.