What we do not measure, and why

Most tools in this market imply they can see inside Google. They cannot, and neither can we. The honest position is to measure the on page signals a machine reads directly, name the off page ones we can only estimate, and be plain about the signals that are real but locked inside Google where no external tool can measure them. This page is that third list.

The 2024 Content Warehouse leak and the DOJ trial confirmed a set of signals Google records per document or per site. They are real. They are also internal, computed from data only Google holds, so a claimed number for any of them from an outside tool is a guess dressed as a measurement. We do not fabricate them.

Signals we deliberately do not fake

• Site authority. Google stores a per site siteAuthority signal, the very "domain authority" it denied for years. It is real and internal. The domain authority and domain rating scores other vendors sell are their own link graph predictions, not this, and not a Google ranking factor. We show your internal link equity distribution instead, which we can actually compute from your own crawl, and we call it that, never "authority".
• Clicks and NavBoost. Google uses click signals, confirmed under oath, held as fields like good clicks, bad clicks and last longest clicks. Only Google sees them. No external tool has your real click stream, so anyone scoring your "user signals" is estimating.
• Chrome telemetry. Aggregate behaviour from Chrome feeds some signals. It is Google's data, not yours to read from outside.
• Sandbox and site age demotions. A new site can be held back for a period. The demotion effect is internal. Site age itself is lookupable, but the effect on ranking is not measurable from outside.
• Panda and quality demotions. Site wide quality demotions exist as internal classifiers. Their codenames leaked, their inputs did not.
• Rater labels and quality scores. Human quality rater labels and internal quality predictions such as the ones nicknamed in the leak are training data and model outputs Google holds. E-E-A-T lives here too: it is a rater target and a set of signals, not a score you can read off a page. See trust and E-E-A-T for what we can honestly check.
• Small site and spam classifiers. Flags such as a small personal site classifier, and the spam models, are internal judgements about your whole site that no crawl can reproduce.

Why publish this at all. Because a tool that pretends to measure the unmeasurable trains you to chase a number that is not real. We would rather be the tool you can trust on the things that are: the deterministic on page checks that make up machine readiness, which have ground truth, paired with clearly labelled estimates for the off page half. If a signal is Google's alone, we say so, and we leave it off the score.