Domain Authority is a vendor's score, not Google's

Who invented the score, and what it was built to do

Domain Authority is Moz's. It's a 1–100 score produced by a machine-learning model trained on Moz's own link index, and its stated purpose is comparative: to predict how likely one domain is to rank relative to another. Moz has been consistent and public about the fact that Google doesn't use it.

Domain Rating is Ahrefs'. It's a 0–100 score computed from the backlink graph alone — how many unique domains link to you, weighted by their own DR, and divided across how many domains each of them links out to. It says nothing about your content, your relevance or your traffic. A site can hold a DR of 70 and rank for nothing.

Semrush has Authority Score, which blends link signals with estimated organic traffic and spam indicators. Majestic has Trust Flow and Citation Flow, which do something different again.

All four are useful instruments built by serious companies, and none of them is what you think it is when it appears on slide two of a monthly report. They're estimates of a proxy, computed from a partial crawl of the web, calibrated against each vendor's own index. They were built for comparing prospects at scale, not for grading an account.

Why your score moves in a month when your site didn't change

This is the property that makes the number unusable as a progress metric, and almost nobody explains it in a client report.

Both DA and DR are relative. The scale is anchored to the strongest domains in the vendor's index, and everything else is placed underneath on a curve. If large sites gain links faster than you — and they always do — you slide down without losing a thing. You are being graded on a curve set by Wikipedia and Amazon.

They're also logarithmic. Going from 20 to 30 might take a few dozen decent links. Going from 70 to 80 can take thousands. So a report showing "DA up 4 points this quarter" means something entirely different at the bottom of the range than at the top, and nobody's report says which.

Then there's recalibration. Vendors re-train these models periodically, and when they do, scores shift across the whole web overnight. Moz has done this openly more than once. Your site got better or worse by four points while you were asleep, on a Tuesday, for reasons internal to a company in Seattle.

And finally, index coverage. The vendor only knows about links its own crawler found. Two tools will report different referring-domain counts for the same site on the same day, sometimes by a wide margin, because they crawled different portions of the web.

• A single site-wide link inflates it. One footer link from a large domain replicated across 40,000 of its pages is one referring domain, but the weighting can move a small site's score noticeably.
• Links to pages that no longer exist keep counting until the vendor recrawls, which can take months.
• Disavowing does nothing to it. Google's disavow file is invisible to third-party tools, so a cleaned-up profile still scores its old toxic links.
• The score can rise while traffic falls. Nothing in DR knows whether your pages match search intent, and intent mismatch is a far more common cause of flat rankings than link shortage.

The circular bit: link sellers price by DR

Here's where a harmless estimate turns into a market distortion.

Go to any link marketplace and the inventory is priced by the host site's third-party authority score. DR 30 costs a few thousand rupees. DR 60 costs several times that. DR 80 on something that looks like a real publication costs six figures. The score is the SKU.

Which creates an obvious business opportunity: manufacture the score cheaply and sell links off it. Because DR is computed from links alone, it can be raised without traffic, readers or editorial standards. Buy an expired domain that still has its old backlink profile, rebuild something on it, point a network of other sites at it, and you have DR 60 inventory in a few months. The site has no audience. It has a price.

So the metric prices the market for links, and the market for links exists to move the metric. You can raise DR by buying from people who raised their DR by buying. Nothing in that loop involves a customer, and Google's link spam systems have spent a decade learning to recognise exactly this shape.

This is the practical reason we don't use authority scores to choose link targets, and why paid links and earned links aren't two flavours of the same thing. The question worth asking about a prospective link is whether the page it sits on gets read by anyone who might buy from you.

Why agencies keep putting it on slide two

We should be fair about the commercial logic here, because it isn't only cynicism — it's the structure of monthly reporting under a slow-moving discipline.

SEO takes three to six months to produce anything a founder can feel. But the report is due every thirty days. And the agency needs something on that report that moved, because "nothing has changed yet, which is expected" is a true sentence that has ended a lot of retainers.

Authority scores are perfect for that job. They tick upward with routine link activity. They're a single number, so they need no explanation. They go on a line chart. They're produced by a tool the client can log into and verify. And crucially, they're not the client's revenue, so nobody has to explain why revenue didn't move.

The result is a report that is honest in every particular and misleading as a whole. The links were built. The score did rise. And the business is exactly where it was.

1. It moves monthly, which is the only real requirement of a metric on a monthly report.
2. It's the agency's own output, since links are the deliverable, so the number measures effort rather than outcome.
3. It's comparable to competitors, which produces a satisfying slide with three bars on it, none of which correspond to anyone's sales.
4. Nobody can be blamed for it. If DA is up and leads are flat, the frame has already been set: the SEO is working, the market is difficult.

Three numbers to report instead

Every one of these can be produced monthly, verified by you independently, and connected to money. Between them they cover the same ground authority scores pretend to cover, without the vendor in the middle.

When a third-party score is genuinely useful

We'd be overcorrecting if we told you to ignore these numbers entirely. We look at them. They're the right instrument for a narrow set of jobs, all of which share one feature: you're comparing many unknown sites quickly, and you don't need precision.

Used this way, the score is a coarse filter — a first pass before a human looks. That's what it was built for.

• Triage at volume. Screening 500 outreach prospects down to 80 worth reading. A rough score is a reasonable first cut when the alternative is reading 500 sites.
• Anomaly detection. A DR 70 domain with almost no organic traffic is a strong signal that the profile was manufactured. Here the score is useful precisely because it disagrees with reality.
• Relative comparison at one moment. Comparing two sites in the same index on the same day is the only comparison these scores were designed to support. Comparing your own score across a recalibration is not.
• Estimating a competitor you can't see inside. You'll never have a rival's analytics. A link-graph estimate is better than nothing when you're sizing up how hard a market will be.
• Sanity-checking an inherited profile. A sudden historical spike in referring domains usually means somebody bought links with your domain. The tools are good at surfacing that shape.

What to ask at your next monthly call

If your report currently opens with an authority score, you don't need to fire anyone. You need to change what's on page one, and a decent agency will do that without complaint — plenty of them report DA only because no client has ever asked for something better.

Four sentences will do it.

1. "What was our qualified organic lead count in the 90 days before we started, and what is it now?" If nobody wrote the first number down, that's the real finding, and it's fixable this month.
2. "Can we agree a list of commercial queries and report position coverage on it every month?" Agree the list together, freeze it, and only change it deliberately.
3. "Of the links built last quarter, how many host sites receive organic traffic, and how many sent us a click?" The gap between links built and links that exist as far as anyone else is concerned is often instructive.
4. "Move DA to the appendix." Not deleted — it's a fine diagnostic and it costs nothing to keep. Just not the headline, because a number that can rise all year while the business gets nothing has no business leading a report. How to read an SEO report covers the rest of the running order.