A site can be internally coherent and still be interpreted through contradictory off-site signals. Misalignment is not noise around the canon; it is part of the interpretive field the model actually reads.

Operational definition

Interpretive misalignment is the gap between the canonical on-site representation of an entity and the off-site signals that a generative system may arbitrate against it. The purpose of the map is not to eliminate every contradiction, but to decide which contradictions alter authority, perimeter, or identity.

Why misalignment must be mapped instead of merely “cleaned up”

Off-site surfaces do not have the same authority, editability, or persistence. Some contradictions are low-value drift; others reweight the source hierarchy or reactivate obsolete interpretations. A map is needed to sort contradictions by structural impact rather than by volume alone.

Types of misalignment

  • Identity misalignment: names, roles, or entity boundaries diverge across surfaces.
  • Offer misalignment: services, capabilities, or exclusions are reconstructed differently off-site.
  • Authority misalignment: third-party surfaces override the on-site canon by frequency or apparent credibility.
  • Temporal misalignment: outdated off-site versions remain active in the synthesis.
  • Normative misalignment: tentative or contextual statements are reframed as stable rules.

Operational method

  • Start by classifying contradictions by affected layer, not by publisher type alone.
  • Distinguish editable from non-editable surfaces and current from archival ones.
  • Prioritize contradictions that change identity, perimeter, or source hierarchy.
  • Correct the canon first when ambiguity exists on-site, then arbitrate off-site where possible.
  • Use canonical cross-references and versioning to stabilize the preferred reading.

What this map prevents

  • Assuming that a perfect site automatically defeats contradictory external signals.
  • Treating all contradictions as equal when only some are interpretively dangerous.
  • Correcting surfaces symmetrically without an authority model.
  • Leaving outdated off-site claims active because they are not technically “wrong enough”.