View packages
A tall mountain peak glowing at sunrise above smaller hills and misty valleys

How is domain authority calculated? Moz feeds a site’s link profile — linking root domains, total links, and spam signals from its own web index — into a machine learning model trained on real Google search results. The model outputs a 1–100 score on a logarithmic scale: a prediction of ranking likelihood, not a Google metric.

On this page
  1. The Calculation Runs on Moz’s Link Index, Not Google’s Data
  2. The Core Input: Linking Root Domains
  3. The Machine Learning Model Behind the Score
  4. What the Calculation Ignores
  5. Why the Scale Is Logarithmic, Not Linear
  6. Why Your DA Moves When You Changed Nothing
  7. Domain Authority vs Page Authority: Same Engine, Different Lens
  8. How Often Moz Recalculates Domain Authority
  9. Domain Authority Calculation FAQ

This guide opens up that calculation step by step: the link index it runs on, the signals that matter most, the machine learning rebuild that changed everything, why the scale is logarithmic, and why your score can move when you changed nothing. If you want the plain-English definition first, read what domain authority means — here we stay inside the machinery.

A minimalist vertical gauge with red fill rising through the lower half against a light background

The first thing to understand about how domain authority is calculated is what it is calculated from. DA is not computed from Google’s data — Google shares nothing. It is computed from Moz’s own web crawl, the Link Explorer index, which discovers pages and links the way any search engine crawler does, just at a smaller scale.

This has a practical consequence most people miss: your DA only reflects the links Moz has found. If a site links to you but Moz has not crawled that page yet, that link is invisible to your DA. It also means DA can lag behind reality after you earn new links, and it can differ from similar scores (like Ahrefs’ DR) simply because each company crawls a different slice of the web. Moz documents the metric’s foundations on its Domain Authority guide, and the index behind it is browsable in Link Explorer.

The Core Input: Linking Root Domains

Inside that index, the heaviest-weighted signal is linking root domains — the number of distinct websites linking to you — rather than the raw count of links. One hundred links from a single site count far less than one hundred links from one hundred different sites, because the model treats each new referring domain as an independent vote of confidence.

Total link counts, follow-versus-nofollow mix, and the authority of the linking pages all feed in as secondary signals, but they orbit the root-domain count. This is also why the calculation discounts manipulative patterns: a sudden burst of links from a cluster of low-quality domains looks, to the model, like the link schemes it was trained to recognize — which brings us to the machine learning layer.

To see why root domains dominate, consider two sites. Site X has 500 backlinks but they come from only 5 domains — a few directories and partner pages. Site Y has 60 backlinks from 50 different domains, mostly editorial mentions. The model scores Site Y higher nearly every time, because fifty independent editorial decisions to link are fifty separate pieces of evidence, while five hundred links from five webmasters are five. When you look at your own link profile, the “linking domains” number tells you far more than the “total links” number ever will, so make it the first metric you check.

The Machine Learning Model Behind the Score

Domain Authority was rebuilt from the ground up in early 2019 — the update Moz called DA 2.0 — and the rebuild is the reason the modern calculation works the way it does. Instead of a fixed formula, Moz trained a machine learning model on actual search results: the model studies the link profiles of sites that really do rank and learns which link patterns predict ranking ability, then scores every domain against what it learned.

Two things changed with that rebuild. First, the score became a genuine prediction — it answers “given this link profile, how likely is this site to rank?” rather than “how many links does this site have?” Second, spam detection got dramatically better, because the model learned what manipulated link profiles look like from real examples. Link networks, paid-link footprints, and other artificial patterns get identified and discounted automatically, which is why spammy link building stopped moving DA the way it once did.

Under the hood, the model weighs far more than domain counts. It reads anchor text distribution (natural profiles use varied anchors; manipulated ones repeat exact-match phrases), link velocity (a sudden spike of new links looks different from steady accumulation), the topical relevance of linking domains, and the authority of the specific pages doing the linking. No single one of these decides your score — the model finds the combination that best matched real ranking outcomes in its training data, which is exactly what “machine learning” means here: correlation discovered from SERPs, not a formula a human wrote.

What the Calculation Ignores

Just as revealing is what never enters the calculation. The model never reads your content — word count, “quality,” and on-page optimization are invisible to it. Domain age is not a direct input either; older domains tend to score higher only because they have had more years to accumulate links. Social shares, brand searches, and traffic are absent too, because none of them live in a link index. Moz cannot measure what it cannot crawl.

This boundary explains two common confusions. First, why a site with excellent content can have a low DA: the calculation scores the link graph, not the writing. Second, why DA sometimes disagrees with your own sense of a site’s strength: your sense includes things like reputation and traffic that the model literally cannot see. DA is a link-profile prediction, nothing more — and understanding its blind spots is part of understanding the number.

Why the Scale Is Logarithmic, Not Linear

The model’s raw output is stretched onto a 1–100 logarithmic scale, and the logarithm is doing real work. On a log scale, each step up requires exponentially more link authority than the last — climbing from 20 to 30 is routine, while climbing from 70 to 80 takes an enormous link profile. That mirrors reality: the gap between an average blog and a mid-tier publisher is small compared with the gap between a mid-tier publisher and a site like Wikipedia.

This is also why DA is best read comparatively. A DA of 35 means little on its own; a DA of 35 against competitors averaging 28 means a lot. For a side-by-side with the other major third-party score, see our DA vs DR comparison — the two metrics are calculated from different indexes with different models, so small disagreements between them are normal.

Overlapping paper bar charts in green and beige tones on a wooden desk with a ruler and pencil

Why Your DA Moves When You Changed Nothing

The most confusing part of the calculation is that it is re-run, not continuous. Every time Moz refreshes its link index — roughly monthly — every domain’s score is recalculated from the new crawl data. And because the scale is relative, your score can fall even when your link profile improved: if the sites around you gained more, or if the index simply crawled deeper into your competitors’ backlinks this cycle, the model re-fits the curve and your number shifts.

Overnight drops almost always trace back to one of three mechanical causes: an index update recrawled your link profile and found lost or devalued links, the logarithmic curve re-fitted around stronger competitors, or Moz improved its spam detection and discounted links it previously counted. None of these mean Google changed its opinion of you — DA is not a Google signal, and Google has repeatedly said it uses nothing like it. A single-index wobble is noise; a three-month slide is a trend worth investigating.

“Lost links” in this context has a precise mechanical meaning: the linking page was deleted, the link itself was removed, or Moz’s crawler simply could not reach the page this cycle. A link that exists on the live web but was not in the latest crawl does not feed the calculation — another reminder that DA measures Moz’s index, not the internet.

Domain Authority vs Page Authority: Same Engine, Different Lens

Moz runs the same style of calculation at two scopes. Domain Authority scores the entire domain’s link profile; Page Authority applies a similar machine learning model to a single URL’s link profile. A page on a DA-80 domain can still have low PA if nothing links to that specific page — the domain’s strength does not automatically flow downhill.

In practice, DA is the number people quote when comparing sites (for outreach targeting or competitive analysis), while PA is more useful when evaluating a single page’s ranking potential. Both are predictions from Moz’s models, both run on the same index, and both shift on the same recalculation cycle.

How Often Moz Recalculates Domain Authority

There is no live, real-time DA. Scores update when Moz pushes a new link index, which happens roughly once a month — and the rollout itself can take days, so different checkers may briefly disagree. If you earned strong links last week and your DA has not moved, the likeliest explanation is simply that the next index refresh has not happened yet, or that Moz has not crawled the linking pages. You can watch your own score move through our free authority checker after each update cycle.

Domain Authority Calculation FAQ

Is Domain Authority a Google ranking factor?

No. Domain Authority is calculated entirely by Moz from its own link index — Google does not use it, see it, or share data with it. It correlates with rankings because strong link profiles tend to rank well, not because Google consults the score.

Why did my DA drop overnight?

Almost always an index refresh: Moz recrawled, found lost or devalued links, or re-fitted the logarithmic scale around stronger competitors. Check whether the drop persists across two index updates before treating it as a real problem.

Do nofollow links count toward Domain Authority?

They can appear in Moz’s index and feed the model as part of the overall link profile, but they carry far less weight than followed links from the same domains. The model is trained on what predicts rankings, and followed editorial links predict rankings best.

How often is Domain Authority updated?

Roughly monthly, when Moz releases a new link index. There is no fixed public calendar, and rollouts take days — brief disagreements between DA checkers during a rollout are normal. A practical rhythm: check your DA once a month right after Moz announces an index update, and log it next to your linking root domain count. Comparing DA without the domain count beside it hides whether a move came from your links or from the curve re-fitting around you.

Can I calculate Domain Authority myself?

Not exactly — the machine learning model and its training data are proprietary to Moz. But you can approximate the inputs: count your linking root domains in Moz’s free Link Explorer, assess their quality, and compare against competitors. That comparison is what DA is for anyway.

Done-for-you podcast backlinks

Get your brand on Apple Podcasts, Spotify, Amazon and 25 more trusted platforms — placed by hand, reported link by link.