What each score is actually measuring
Both numbers try to answer "how strong is this domain?", which isn't a question with a real-world unit attached. Each vendor answers it differently, from a link index only they can see.
Domain Rating (Ahrefs) is a pure link-graph calculation: count the unique domains linking to you, weight each by its own DR, divide that weight across how many domains it links out to. It knows nothing about your content, traffic or rankings. It describes a graph.
Domain Authority (Moz) is a machine-learning model trained to predict how often a domain appears in search results, built on Moz's own index. It takes more inputs than raw links and it's a *relative* score, calibrated against the rest of the web — so your DA can fall while your links grow, simply because bigger sites grew faster.
| DA (Moz) | DR (Ahrefs) | |
|---|---|---|
| Owner | Moz | Ahrefs |
| Scale | 1–100, logarithmic | 0–100, logarithmic |
| What it models | Predicted likelihood of ranking | Backlink profile strength only |
| Inputs | Link index plus model features, calibrated against the web | Referring domains, weighted by their DR and outbound domain count |
| What it ignores | It's a prediction, not a measurement | Content, traffic, rankings, relevance, spam |
| Recalibrated | Yes — as the web grows | No |
| Moves when | Moz recrawls or retrains | Ahrefs finds or drops referring domains |
| Ease of manipulation | Harder — spam-shaped profiles predict badly | Easier — redirect or buy enough weighted domains |
| Used by Google | No | No |
A worked example: DA 21, DR 61, same site
This is a constructed illustration, not a client — but it's the exact shape we see when a founder forwards us a link-selling deck and asks whether it's a good buy.
A five-year-old site in a commercial category. 340 referring domains. DR 61. DA 21. Forty points of disagreement about the same domain.
| Source of links | Domains | Effect on DR | Effect on DA |
|---|---|---|---|
| Three link-network footprints — generic names, no named authors, hundreds of outbound links each, three repeated anchors | 120 | Large. They carry inherited DR and DR doesn't judge quality. | Small or negative. A model tuned on ranking has learned this shape doesn't. |
| Profile and directory pages on high-DR platforms | 60 | Real — the host domain does the work. | Near zero. These pages rank for nothing. |
| One expired domain, 301-redirected into the homepage in 2023, carrying its own old profile | 45 | Immediate and large — a redirect folds another domain's graph into yours. | Minimal. Moz's model isn't crediting the graft. |
| Genuine editorial links from small, real, relevant sites | 115 | Modest. Small sites have little DR to pass. | This is the pile DA rewards — and 115 of them is a DA 21 site. |
So which one is lying?
Both, a little, in different directions.
DR describes a link graph faithfully. The lie is one of implication: a high DR reads as "strong site" when it may only mean "a lot of weighted domains point here, by whatever means". DR has no opinion about how they arrived.
DA is trying to predict something real, which makes it more useful and less literal. But it's a model, it runs on a smaller index, and it's relative — so a DA drop can mean the web grew, not that you shrank. Founders panic about that every quarter and there's usually nothing to fix.
The deeper problem: both answer a question you don't have. You don't need a domain's abstract strength. You need to know whether a link from that page will help *your* page rank for *your* query, and neither score is built for that. Longer case in domain authority is a vanity metric.
Neither number is Google's
This part isn't ambiguous. DA is a Moz product, DR is an Ahrefs product, both built by third parties reverse-engineering an index they can't see. Google has said repeatedly that it doesn't use Moz's Domain Authority.
One nuance, stated honestly rather than pretended away: internal Google documentation that surfaced publicly in 2024 referenced site-level quality signals, and SEOs have argued about it since. It doesn't imply Google reads DA or DR — those are vendor scores from vendor indexes.
The practical consequence: any contract that sets a DA or DR *target* has made a third party's opinion your deliverable. If Moz retrains its model in March, the KPI moves without anybody doing any work. That's a weather report, not a metric.
How link vendors move these numbers on purpose
None of what follows is secret. It's the standard toolkit of the Indian and global link marketplaces, and knowing it is enough to disqualify most of what lands in your inbox.
- 301 stacking. Buy an expired domain with an old link profile, redirect it at the money site. DR jumps within weeks because the redirect folds an entire graph in. Nothing about the site got better.
- Expired-domain rebuilds. Same trick without the redirect: resurrect a dead domain on its old URL structure, publish thin content, sell placements on the residual authority. The tell is a traffic curve that's flat for years, dies, then restarts.
- Sitewide footer and blogroll links. Fifty links, one referring domain. Sold as "50 backlinks", worth roughly one.
- Link exchange rings. Three- and four-way swaps to dodge the obvious reciprocal footprint. Whole marketplaces run on it.
- Platform profile spam. Mass profile pages on genuinely strong platforms. Strong host, page nobody has ever crawled with interest.
- Quoting whichever score wins. The same inventory gets pitched as "DA 45" in one deck and "DR 72" in another. If a seller only shows one of the two, ask for the other and watch the pause.
What to check instead, before you pay for a link
Eight checks, about two minutes a site once you've done a few. They'll disqualify most paid inventory, which is the point — see paid links versus earned links for what the alternative costs.
- Organic traffic to the linking page, not the domain. You want "some" versus "zero"; estimator precision doesn't matter.
- The traffic curve over three years. Flat, cliff, restart means a rebuilt expired domain.
- Does the site rank for its own brand name? If not, Google isn't confident it exists.
- The outbound-link pattern on the page. Forty links to unrelated commercial sites is a farm, whatever the score says.
- Topical relevance to your page, not your industry generally.
- Is the exact page indexed? Paste a unique sentence into Google in quotes. Unindexed page, worthless link.
- Is there a named human? Author page, real LinkedIn profile, masthead, address. Networks skip this because it costs money.
- Would you send a paying customer to this page? The most reliable filter in link buying, and it takes a second.
The verdict
Use DR for triage. It's fast, directionally useful for sorting 200 prospects down to 40, and everyone in outreach already speaks it. Never use it as the decision.
Use DA for almost nothing you're paying for. It's a more thoughtful number, but it's slower to update, built on a smaller index, and relative in a way that produces quarterly false alarms.
Use traffic and relevance as the actual test. A page with 200 real monthly visits, a named author and a topic adjacent to yours beats a DR 70 host with no traffic every time — and costs less, because the marketplaces haven't priced it.