<b>Two jobs one anchor does: relevance and trust are separate axes</b>

<b>Two jobs one anchor does: relevance and trust are separate axes</b>

The over-optimization conversation conflates two functions an anchor serves, and the conflation hides where the actual risk lives. An anchor simultaneously transmits a relevance signal — what the target is about — and participates in a trust assessment — whether the link looks earned or manufactured.

These axes can move independently, which the bucket model obscures. A perfectly relevant exact-match anchor maximizes the relevance signal while minimizing the trust signal if it appears in a manipulative pattern. A vague generic anchor does the reverse.

— Relevance asks: does this anchor help the engine understand the target?
— Trust asks: does this anchor's pattern look like something a human editor would actually write?

On one hand, this explains why a few strong exact-match anchors from genuine editorial contexts seem to help rather than harm — they score high on both axes at once. On the other, it explains why mass exact-match fails: it maxes relevance while collapsing trust.

Limitation: I am describing a conceptual decomposition, not a measured two-dimensional model; the engine does not publish separate scores, so the split is interpretive.

Open question: is the durable strategy simply to maximize relevance only on links that already score high on trust — and to never let the relevance dial move the trust dial?