What is Domain Authority and Does It Actually Matter?
A third party estimate that became the industry's default currency, despite not being used by any search engine. It is neither useless nor what most buyers think it is. Using it well starts with knowing exactly what it measures.
What Domain Authority Is
Domain Authority is a score, usually on a scale of one to a hundred, produced by a commercial SEO tool vendor to estimate how likely a website is to rank in search results. Several vendors publish their own version under their own names, which is why the phrase gets used loosely for all of them.
The most important fact about it belongs in the first block rather than buried later. It is not Google's. Google does not publish a score of this kind, does not calculate one for public use and has no involvement in any vendor's figure. What you are looking at is a private company's estimate of how another company's systems might behave.
One property worth knowing early, because it explains a lot of frustration. Scores of this type are not linear. Moving a site from twenty to thirty is achievable. Moving from seventy to eighty is a different order of difficulty entirely, since the upper end of the scale is occupied by a small number of very large sites and everything else is compressed below them.
What follows covers where the number came from, what it is genuinely useful for, where it misleads and what to ask when somebody quotes one at you. The full series sits in the complete guide to backlink building.
Where It Came From And Why It Spread
The origin is commercial rather than scientific. Understanding that explains most of what follows.
Search visibility is difficult to reason about. Dozens of factors, none published, all interacting, with outcomes that shift. Buyers needed something they could compare, budget against and put in a report. Vendors, sensibly, built a single figure that summarised what their crawlers could see.
What happened next was not designed by anybody. The number became the unit of trade. Sellers began pricing placements by it. Buyers began specifying it in briefs. Reports began leading with it. Whole marketplaces organised themselves around bands of it, which is the subject of our guide to DA90 backlinks.
The point worth sitting with is why it spread. Not because it turned out to be accurate. Because it was convenient, comparable and easy to put on a slide, which are properties of a good unit of account rather than properties of a good measurement.
And once a number becomes a currency, something else follows. There is now commercial pressure on that number which has nothing to do with its accuracy, because a whole market's pricing depends on it. Any measure that becomes a target starts behaving differently from a measure that is only observed.
How Scores Like This Are Built
Broadly, from link data the vendor has crawled itself. It looks at which sites point at which other sites, applies a model and produces a figure. The exact method is proprietary in every case, which is itself worth noting: you are asked to trust a number whose derivation is not disclosed.
Four consequences follow from that construction. Each explains a behaviour buyers find puzzling.
It lags. The score reflects what that vendor's crawler last saw. A site that has changed recently, in either direction, is described by information of uncertain age.
It varies between vendors. Different crawls, different models, different scales. Two vendors will give the same site different numbers. Neither is wrong, because they are not measuring the same construct. Comparing a figure from one against a figure from another is meaningless.
It is a prediction rather than a measurement. These scores estimate likelihood of ranking. A prediction is a model output that can simply be wrong about any individual case, in a way that a measurement of something real cannot.
It is usually about the domain. Which discards everything that matters about the specific page carrying a link, as set out in what relevance really means in backlink evaluation.
Why Google Does Not Use It
Google does not use any third party authority score. It could not, even if it wanted to. These figures are calculated by outside companies from their own crawls, arriving at conclusions about Google using none of Google's data.
Google's search representatives have said publicly and repeatedly that there is no single site wide authority score of this kind operating inside their systems. They have also discouraged treating vendor metrics as though they described Google's behaviour.
The relationship is worth stating precisely, because the useful version is more subtle than a flat dismissal. A vendor's score is an outsider's attempt to predict an outcome, built by a company with no access to the process producing it. Sometimes such predictions correlate with reality, in the sense that sites which rank well tend to have the properties these models reward. That correlation is why the numbers are not worthless.
What the correlation does not license is treating the estimate as the thing itself. A weather forecast has genuine value while remaining a forecast. Nobody argues with the sky on the grounds that the forecast said otherwise.
So the reasonable position is neither that these scores are meaningless nor that they measure authority. They are one company's estimate, useful within limits. Those limits are the subject of the next three blocks.
What It Is Actually Useful For
The score has real uses. Pretending otherwise would leave a reader without a practical alternative to the thing they were already doing.
Triage at volume. Given five hundred candidate sites and limited time, sorting by a score tells you where to start looking. It is a queue rather than a verdict, which is a legitimate job for a rough number.
Spotting outliers. A site presenting itself as an established publication while carrying a very low score is worth examining. So is the reverse: an unremarkable site with an unexpectedly high one, which frequently indicates the score was worked on deliberately.
Rough comparison within a field. Two competitors in the same sector, measured by the same vendor on the same day, produce a comparison that means something approximate. Not precise, though directionally informative.
Very long term tracking of your own site. Watched over years rather than months, using one vendor consistently, the trend carries some information about whether a profile is growing.
What unites all four is that the score is being used to decide where to look rather than what to conclude. Used that way it saves time. Used as an answer it substitutes for the judgement it was supposed to precede.
Where It Misleads
Three failure modes. The first one supports an entire market.
The score can be worked on. Any figure derived from links can be moved by acquiring links. A site can therefore be built specifically to score well, without ever acquiring readers, standing or a reason to exist. That is not a hypothetical: it is the business model behind placements sold by score band. It means a high number is entirely compatible with a worthless site.
It contains no relevance. The most consequential omission. A score describes a domain in isolation, saying nothing about whether that domain has any connection to your field. A modest specialist publication in your own trade is worth more than a large unrelated site. The score reverses that ranking every time.
It describes a domain while links live on pages. A strong publication can host a page nothing has indexed, that nobody reads, in a section abandoned years ago. The domain score is unaffected. The link is worth nothing.
Underlying all three is a familiar problem. Once a number becomes the specification, suppliers optimise for the number rather than for the outcome it was standing in for. Anybody buying by score is, in effect, commissioning work against the proxy instead of against the objective.
What To Use Instead
Six checks, all free, taking a few minutes per candidate. They replace the score with the thing the score was approximating.
Is the specific page relevant? Not the site. The page carrying the link, plus the paragraph around it.
Is that page indexed? One search on a distinctive sentence settles it. A page nothing has recorded cannot contribute anything at all.
Does anybody read it? Is there any route by which a human being would arrive there other than a link report.
Was the link editorial? Did somebody with responsibility for the page decide to include it, in service of readers.
Would you want to be seen there? Judged as a customer would judge it, with no thought of rankings.
Does the query you care about even turn on links? For local intent it largely does not, whatever the source scores.
A site answering yes to those is worth pursuing whatever its number. A site answering no to several is not worth pursuing whatever its number. The full reasoning behind each check is in what relevance really means in backlink evaluation.
If Somebody Quotes You A Score
Whether in a proposal, a report or a sales call, four questions establish what you have actually been told.
Whose metric is it? Different vendors publish different scores under different names, on different scales. A supplier who cannot say which one they are quoting is repeating a figure rather than using it.
When was it measured? These figures are recalculated periodically and can move without anything changing on the site itself. A number in a proposal has a date, even when the proposal does not show one.
Is it for the domain or the page? Almost always the domain. Ask anyway, because the answer determines whether the figure has any bearing on the link being discussed.
May I see the page? The one that ends most conversations. It is also the only one that cannot be answered with another number. A supplier confident in a placement will show you the page immediately. Hesitation here tells you more than any of the previous three answers.
The wider set of questions to put to any supplier is in questions to ask before buying backlink services.
We do not report
a score as a result.
Every reference we obtain is reported as a live address you can open and judge for yourself. We use vendor metrics to decide where to look and never to decide what to conclude. No purchased placements, no networks, no price per link.
Included every month:
One fixed monthly rate. No setup fee. No price per link.
Twenty four guides.
One subject.
This page covers the metric. The rest of the series covers what it stands in for, what gets sold on the strength of it, how to audit a profile properly and when links are not the right spend at all.
