What is Knowledge-Based Trust?

Knowledge-Based Trust (KBT) is a computational method developed by Google Research that scores the trustworthiness of web sources based on the factual accuracy of their content — not on how many links point to them.

The Core Idea

Traditional web ranking, epitomised by PageRank, infers quality from popularity: a page linked to by many authoritative pages is assumed to be good. KBT inverts this logic. Instead of asking "who links here?", it asks "are the facts here correct?"

A source accumulates a high KBT score by making claims that are consistently supported by a trusted knowledge base — and loses score when its claims contradict established facts. Popularity without accuracy yields no credit.

How KBT Works

The method, introduced by Dong et al. in their 2015 VLDB paper (arXiv:1502.03519), proceeds in three steps:

1. Fact extraction. An information-extraction pipeline pulls subject–predicate–object triples (e.g., "Albert Einstein — born in — Ulm") from web pages at scale.
2. Cross-checking against the Knowledge Vault. Each extracted fact is compared to Google's large-scale knowledge base. A fact is labelled correct, incorrect, or uncertain based on that comparison.
3. Score aggregation. The proportion of correct-to-incorrect facts across all claims on a source yields its KBT score — a value between 0 and 1.

Endogenous vs. Exogenous Trust

KBT is an endogenous trust signal: it derives entirely from the content of the source itself, without reference to external links or citations. This is what makes it structurally different from PageRank and domain-authority metrics, which are exogenous — they depend on what other sources say about the source. See Endogenous vs. Exogenous Trust for a deeper comparison.

Why It Matters for SEO and AI Visibility

Search engines and AI answer engines increasingly weight factual accuracy as a ranking signal. Pages that make verifiable, accurate statements about well-established topics are more likely to be cited by large language models and to surface in AI-generated answers. KBT provides the theoretical backbone for understanding why factual correctness — not just topical relevance — is becoming a primary quality criterion. For the relationship to Google's public guidelines, see Knowledge-Based Trust and E-E-A-T.

About This Site

This resource is maintained by Johannes Faupel, Frankfurt am Main, as an educational reference on KBT and its implications for web publishing and AI citation. All articles cite the primary research literature.