A playbook for auditing your page against Google's entity extraction

A playbook for auditing your page against Google's entity extraction

Question: which entities does Google actually parse from your page, versus the ones you intend? Run this before any topical-authority build.

— Step 1: paste your raw body text into the Cloud Natural Language demo (the public one Google ships). Note every entity it returns with a salience score above 0.05.
— Step 2: cross-check those against your target topic. Salience is a relative weight (0–1) showing how central an entity is to the document.
— Step 3: list entities you wanted but that scored under 0.01 — these are under-expressed.
— Step 4: add one declarative sentence per missing entity near the top third, then re-run.
— Step 5: confirm your primary entity now sits in the top 3 by salience.

Method note: based on repeated runs of the Cloud NL demo across ~40 ranking pages; salience numbers are Google's own API, not a third-party proxy.

Caveat: the demo isn't the exact production model, and salience does not equal ranking weight — treat it as a directional signal, not ground truth.

Confidence: medium