What the scores are
An authority metric is a single number, conventionally between nought and a hundred, computed by a commercial vendor from its own crawl of the web's link graph. The intuition behind all of them is the same and is genuinely sound: a page linked to by pages that are themselves well linked is probably more consequential than a page linked to by nothing. That recursive idea is old, well understood, and the basis of a great deal of the modern web.
What the scores are not is an input. No search engine publishes an authority figure, and none reads a third party's. Raising a vendor score does not cause a ranking to improve; both may move together when the underlying links change, which is a correlation and is constantly reported as a mechanism. The distinction is not pedantic: it decides whether "improve our score" is a coherent objective or a category error.
The four scores side by side
What each is computed over
| Score | Scope | Built from |
|---|---|---|
| DR: Domain Rating | Whole domain | The strength of the domain's inbound link graph, on a logarithmic scale. Deliberately ignores content and traffic entirely. |
| DA: Domain Authority | Whole domain | A model trained to predict relative ranking ability from link features. Being a prediction, it shifts when the model is retrained, not only when links change. |
| AS: Authority Score | Whole domain | A composite that blends link data with estimated organic traffic and a spam assessment, which makes it less pure and sometimes more useful. |
| PA: Page Authority | Single page | The same modelling as DA applied to one URL. Usually the more relevant of the pair, and much less quoted. |
Two structural differences matter more than the formulas. First, domain-level and page-level scores answer different questions: a strong domain routinely hosts weak pages, and quoting a domain score while discussing a specific page silently substitutes one for the other. Second, one of the four includes traffic in its composite while the others are purely link-derived, so it will disagree with them in a systematic rather than random direction.
Why the scales are logarithmic
All of these scores compress a range that spans many orders of magnitude into a hundred points. The largest sites on the web have inbound link graphs millions of times larger than a modest business site, and a linear mapping would leave every ordinary site clustered indistinguishably near zero. A logarithmic scale spreads the interesting middle out and squeezes the extremes.
The consequence is the one people find counter-intuitive. Moving from twenty to thirty is a modest amount of work; moving from sixty to seventy is an enormous amount, because the underlying quantity required has grown by a large multiple. Equal-looking increments are not equal, and treating the scale as linear turns a reasonable target into an impossible one without anybody noticing.
What none of them can tell you
Every one of these scores is computed from links and nothing else, with one partial exception. So the list of things they cannot see is long and includes most of what actually matters about a site: whether the content is accurate, whether anybody reads it, whether it converts, whether the business behind it is solvent, whether the pages are indexed at all.
Two omissions cause the most trouble in practice. Relevance is absent: a high-scoring domain in an unrelated field is not a better link source than a modest one in the right field, and the score cannot express that. And the nature of the links is flattened: a hundred template footer placements and a hundred considered editorial citations can produce a similar figure, while being entirely different things.
The check the number cannot do for you
Open the strongest few referring pages and read them. Two profiles with the same score, one built from reference citations and one from directory listings, are not comparable assets, and no published metric distinguishes them.
Why they are easier to move than rankings
Because the input is only the link graph, and the link graph can be added to. A ranking has to satisfy relevance, intent and quality assessments as well; a link-derived score has to satisfy arithmetic. This asymmetry is the whole basis of the market in high-scoring domains, and it produces a diagnostic worth knowing: a domain with a high authority score and negligible search visibility is a domain whose links were assembled rather than earned.
Vendors know this and push back with spam scoring, link retirement and periodic recalibration, which is why scores move without any link changing. A recalibration can move an entire market of domains by several points overnight. Any process that treats a score as a stable property of a site rather than a reading from an instrument will be surprised by that at least once a year.
Using a score without misusing it
Four rules cover it. Compare within one vendor only, because the numbers share a range and no methodology. Prefer page-level to domain-level when the subject is a page. Treat the scale as logarithmic when setting any target. And read the profile behind any number you intend to rely on, because the score is a summary of a graph you can actually inspect.
And keep the wider frame in view. These are vendor products, and the history of this layer is that vendor products stop existing: the record of closed measurement tools runs to nine services, including the last free public link index. A score quoted today may be uncheckable in five years, whereas the registration record and the DNS records underneath it will still answer, because publishing them is an obligation rather than a business model.
