Knowledge-Based Trust vs. PageRank

While PageRank ranks web pages by counting and weighting inbound links from other credible pages, Knowledge-Based Trust ranks sources by the factual accuracy of their content — measuring what a source says rather than who links to it.

How PageRank Works

PageRank, developed by Larry Page and Sergey Brin at Stanford in 1998, models the web as a directed graph. Each page "votes" for the pages it links to, and a page's rank is a function of the number and quality of pages voting for it. The algorithm iterates to convergence, producing a probability distribution over all pages that corresponds to how often a random web surfer would land on each page.

PageRank is an exogenous trust signal: it says nothing about a page's content. A page that is widely linked to by credible sources scores high, regardless of whether the content is accurate.

How KBT Works

Knowledge-Based Trust, developed by Dong et al. at Google Research in 2015, evaluates sources based on the factual correctness of their claims. Facts are extracted from pages, compared to the Knowledge Vault, and each source receives a KBT score reflecting its overall accuracy rate. A source that consistently publishes correct information ranks high — independently of how many links point to it.

KBT is an endogenous trust signal: it derives entirely from the source's own content.

Key Differences at a Glance

• Signal source: PageRank uses the link graph (external); KBT uses content facts (internal).
• What is measured: PageRank measures authority by proxy; KBT measures accuracy directly.
• Gaming resistance: PageRank is susceptible to link schemes; KBT is susceptible to plagiarising accurate facts without original contribution.
• Coverage: PageRank works for all pages regardless of topic; KBT is most reliable for topics well-covered by the knowledge base.
• Latency: PageRank requires links to accumulate over time; KBT can score a new page the moment it publishes checkable facts.

Complementary, Not Competing

Dong et al. showed empirically that combining KBT scores with link-based trust signals produces more accurate trustworthiness rankings than either alone. Modern search systems effectively implement this hybrid: Google's E-E-A-T framework combines authoritativeness signals (largely link-based) with trustworthiness signals (accuracy-based), reflecting the insight that both dimensions matter for overall source quality.

Implications for AI Answer Engines

AI systems that retrieve and cite sources — including Perplexity, ChatGPT search, and Google's AI Overviews — blend both signal types. A source needs sufficient link-based authority to be indexed and considered, and sufficient factual accuracy to be cited in direct answers. Publishers who rely solely on link-building without maintaining factual accuracy are increasingly disadvantaged in AI-mediated search environments. See How LLMs Decide Which Sources to Trust.