Shiliu and Tianwang: ad-heavy pages and data security

Two older algorithms that established principles Baidu still enforces through their successors.

Shiliu (石榴) — the low-quality page algorithm

Launched in 2013 and described at the time as “the terminator of low-quality pages”, Shiliu targeted pages overwhelmed by advertising: large volumes of poor adverts, pop-ups, and content buried beneath commercial units. Its stated aim was pages where the user cannot reach what they came for.

Its principles now live in the Mobile Landing Page Experience White Paper and Bingtong, which express the same idea with numbers attached — main content must occupy at least half the first screen, no floating or pop-up advertising anywhere. Shiliu is where that line of enforcement begins.

Tianwang (天网) — data security

Tianwang targeted the theft of user privacy: sites embedding malicious code to harvest visitors’ personal data, particularly identity and payment information. It is a security algorithm rather than a quality one, and the response is correspondingly severe.

Why these two still matter

• Shiliu established that advertising density is a ranking factor in its own right, not a matter of taste. Every subsequent Baidu experience standard builds on that.
• Tianwang established that a site is responsible for what its code does to visitors — including code the owner did not write. Compromised sites and third-party scripts harvesting data fall under this whether or not the owner knew.

The practical reading

For advertising, work to the current specification: the white paper has explicit numbers, and meeting them satisfies both the old principle and the current algorithm. For security, audit what third-party scripts actually collect — analytics, chat widgets, advertising frameworks — because “we did not write it” is not a defence in either the algorithm or the law.

The wider lesson

Baidu’s older algorithms are not retired; they are absorbed. Reading them tells you the direction of travel, and the direction has been consistent for a decade: less interference between the user and the content, and more responsibility on the site for what happens to the visitor.