Domain authority

The confusion has a specific source. Google did once expose a public per-page score, Toolbar PageRank, and for years the industry optimised against a number Google itself published. It was withdrawn in 2016, and vendor scores filled the vacuum it left.

They are not the same kind of thing. Domain Authority, Domain Rating and Authority Score are predictions built by third parties from their own crawls, trained to correlate with rankings. They are estimates of an outcome rather than an input to it, and no search engine reads them.

Google has addressed this directly and more than once, both to say it does not use a domain-wide authority metric and to discourage treating vendor scores as if it did. That is worth having to hand, because clients arrive with the number and usually believe Google assigns it.

The honest framing for a client conversation: it is a third-party estimate of how a domain tends to perform, useful in the way a credit score is useful, and about as connected to the decision being made.

These scores are logarithmic, which is the single most misread property. The distance from 20 to 30 is not the distance from 70 to 80, and treating the scale as linear produces badly wrong expectations about what a campaign can move.

Early movement is cheap. A new domain with a handful of decent links will climb the low numbers quickly, which is why agencies like showing the metric in the first two quarters of an engagement, and why it flatters early work.

It then stalls, and it stalls right when the client has learned to look at it. Explaining in month eight that the number is logarithmic, after eight months of showing it going up, is a conversation that sounds like an excuse regardless of being true.

The other half of reading it correctly is that only one comparison is defensible: a domain against itself, in one tool, over time. Across tools the scores are unrelated, and against a competitor in the same tool they are directional at best, since both numbers are estimates with error bars nobody publishes.

There is one task these scores do well, and it is not reporting. Assessing link opportunities at speed, the score is a reasonable first filter across a list of two hundred sites nobody has time to open.

The tell it surfaces best is the mismatch. A domain with a high authority score and negligible organic traffic is almost always a site built to sell links: the links exist, the audience does not. That pattern is hard to see quickly any other way, and it is the single most useful thing the metric does.

The same applies to prospect research. A quick score on a prospective client tells you roughly what you are walking into before the first call, which is a legitimate internal use.

The rule that follows is simple enough to be a policy: it is an internal triage number. Use it to decide where to spend attention, and do not show it to anyone who will be asked to act on it.

Once a number appears in a client report it is in every report after it, and its movement becomes a standing item whether or not it means anything.

That is expensive for a metric with these properties. It stalls by design, so the chart flattens while the work is going well. It is vendor-specific, so a competing agency can quote a different number for the same domain and appear to contradict you. And it is not consulted by any search engine, so every conversation about it is a conversation about a proxy.

The worst version is when it moves the wrong way. Vendor index updates can drop a score across many domains at once for reasons internal to the tool, and explaining that the fall was a recalculation rather than a loss is a call nobody enjoys.

What to report instead is what the score was standing in for: referring domains earned, and which pages they point at. Those are real events, they are attributable to the work, and they do not stall in month eight.