TrustRank

Long definition

TrustRank was introduced in a 2004 paper by Yahoo Research and Stanford ("Combating Web Spam with TrustRank"). The setup: take a small set of high-trust seed sites curated by humans, then propagate a trust score outward through their outbound links. The further a site sits from the seed in the link graph, the lower its inherited trust. Sites with high trust were treated as credible; sites isolated from the trust graph were candidates for spam classification.

The algorithm itself was Yahoo's. Google has never confirmed running TrustRank under that name, and the trademark was held by Google for a separate purpose then abandoned. But the idea — that link-graph proximity to known-good sites is a usable signal for ranking and spam filtering — appears throughout how Google has described its evaluation systems for two decades. The Quality Rater Guidelines, E-E-A-T, link-based spam classifiers, and the seed-set logic in the Disavow workflow all rest on similar foundations.

Two practical takeaways for SEO today:

• Editorial mentions from trusted hubs compound. A link from a major publication, government site, or established educational domain seeds your site in the trust neighborhood that those hubs anchor. Each step closer to the trusted core lifts the priors of every downstream ranking decision.
• Distance from spam matters too. The inverse of trust is anti-trust. A backlink graph whose nearest neighbors are PBNs and scraper sites pulls your site toward the spam classifier's positive class — even when no individual link is overtly manipulative.

For modern SEO, TrustRank is a useful mental model rather than a dial you can turn. You can't reverse-engineer Google's seed set. But you can ask "if a Google engineer drew a graph of trusted sites and walked outward link by link, in how many hops would they reach me — and what would they pass through?"

Common misconceptions

• "TrustRank is a current Google ranking factor." It's a 2004 academic algorithm from Yahoo Research. Google has never confirmed running it. Modern trust signals share the spirit, not the exact mechanism.
• "You can buy TrustRank with high-DR links." No. The model relies on graph distance from a curated seed, not on third-party domain metrics. A high-DR link from a site disconnected from the trust core does less than a lower-DR link from a site embedded in it.
• "TrustRank and PageRank measure the same thing." PageRank measures link authority neutrally — every link is a vote. TrustRank weights votes by proximity to a trusted seed, explicitly trying to filter spam out. Different goals, different math.