<b>Co-occurrence may matter more than the anchor itself</b>

<b>Co-occurrence may matter more than the anchor itself</b>

Most anchor analysis treats the clickable text as the whole signal. A more interesting hypothesis, traceable to Google's older phrase-based indexing patents (e.g. US 7,536,408), is that the <i>text surrounding</i> the link carries weight independent of the anchor.

The idea: a link with anchor "click here" embedded in a paragraph dense with related entities and co-occurring phrases can transmit topical relevance that the anchor string alone doesn't encode. Conversely, an exact-match anchor parachuted into off-topic boilerplate looks engineered precisely because the surrounding text doesn't corroborate it.

On one hand, this elegantly explains why "click here" links from authoritative, on-topic pages seem to help despite carrying zero keyword in the anchor. On the other, we can't isolate co-occurrence from the dozens of correlated quality signals on a good page, so the causal claim stays soft.

A 2023 anchor-distribution analysis noted that natural profiles carry a long tail of generic and naked anchors that nonetheless sit in topically coherent contexts — which fits the co-occurrence model better than a pure anchor-string model.

Practical implication: judge a prospective link by paragraph context, not just the anchor you can negotiate.

Open question: does modern Google still lean on phrase co-occurrence as a relevance proxy, or has passage-level embedding superseded it — and would the two even diverge in practice?