Domain Authority: What It Is and What Counts as a Good Score

Domain authority is a 1-to-100 score invented by Moz that estimates how likely a website is to rank in Google, based mainly on its backlink profile. It is not a Google metric, Google does not read it, and Moz says both of those things on its own page. Ahrefs, Semrush and Majestic each sell a different score doing a similar job under a different name.

The short answers, before anything else:

• What is a good domain authority score? There is no universal one, and the two companies that sell an authority score both refuse to name a number. The workable answer is relative: good means comparable to the sites already linking to the page you want to outrank. Full answer below.
• How do I raise it? Earn more links from more separate, genuinely editorial domains. There is no other lever that reliably moves any of the four scores. The tactics that actually move it.
• Is it a Google ranking factor? No. Google has never used it. But Google does have site-level signals of its own, and it has named one of them. What Google actually has.
• How accurate is it? Moz published the only coefficient anyone has: a mean per-SERP Spearman correlation with rankings of about 0.12, in its own validation study. That was the best of the four. The number nobody prints.

That is the useful part, and it is all above the fold. The rest of this post is the evidence, which is more interesting than the summary and contradicts a good deal of what ranks for this keyword.

The company that invented domain authority publishes this sentence on the page that defines it: "Domain Authority is not a Google ranking factor and has no effect on the SERPs."

That is Moz, about Moz's own product, in Moz's own documentation. The same page calls the score "a predictive metric developed by Moz" that "is not used by Google to determine search engine rankings." No ambiguity, no small print, no walking it back.

Now read a sentence from Google. It is the opening line of the site reputation section of the published spam policies, and it is describing the abuse it exists to stop:

The site reputation policy applies where third-party content is published on a host site mainly because of that host's already-established ranking signals, which it has earned primarily from its first-party content.

Read that again slowly, because it is doing something nobody quotes it for. Google is saying that a host site has ranking signals. That the site earned them. That it earned them from its own content. And that those signals are transferable enough that renting out a subfolder is a profitable enough scam to need a written policy against it.

So: the vendor says the number is not Google's. Google says the property is real. Both of those are true at the same time, and almost every argument about domain authority is people shouting one of them at someone holding the other.

This post is about the gap between them, which is the only interesting part of this keyword. What Google actually has, what the four vendors actually sell, what the leaked documentation actually calls it, and what the number is honestly good for — which is a real answer, not nothing.

What is a good domain authority score?#

The honest answer is that the question has no fixed answer, and the people best placed to give you one have declined in writing.

Ahrefs, which sells Domain Rating: "It's not possible to say that a Domain Rating of 30 is good, or 50, or 60, or 70. It's all relative." Semrush, which sells Authority Score: "Authority Score is best used for domain comparison, and not for determining good/bad on an absolute scale." Moz, which invented the category: "it's important to use Domain Authority as a comparative metric rather than an absolute one."

Three vendors, three refusals, all published.

The band you have probably seen — 40 to 50 average, 50 to 60 good, above 60 excellent — is real in the sense that plenty of pages print it. I counted where it comes from, and the answer is in the figures below: it appears on three of the ten pages ranking for this keyword, none of which sells an authority metric, and none of which attributes the band to anything.

Use this instead, and it takes four minutes. Open the page you are trying to outrank, export its referring domains, and look at the spread of their scores in whichever single tool you use. That spread is what good means for your keyword. A dentist in Leeds and a developer-tools startup will get completely different answers, which is exactly why no universal number exists.

Two rules of thumb that survive contact with reality:

• Below about 20 in any tool, the score is mostly telling you the site is young or small, not that it is bad. Plenty of genuinely useful trade sites live here, and they are often the easiest links you will ever get.
• Above about 70, the score has stopped discriminating for most purposes. The difference between a 74 and an 81 will not decide whether a link is worth having; whether anyone reads the page will.

How to increase domain authority, and what actually moves it#

Every one of the four scores is, at its core, a model of your backlink profile. So the list of things that move it is short, and anything not on this list is noise.

1. Earn links from domains that have never linked to you before. All four scores count referring domains, not links. Ahrefs is explicit: "Only the first link from a domain will increase your DR. All subsequent links don't affect your DR any further." Ten links from one site is one data point.
2. Prioritise domains that are themselves well linked. Ahrefs describes its calculation as passing "some amount of 'DR juice' from each linking domain," determined roughly "by dividing the DR of the linking domain by the number of unique domains that it links to." A link from a site that links to almost nobody is worth more than one from a site that links to everyone.
3. Make sure the links are followed. "If a website is only linking to you via nofollow links, it won't increase your DR." That is a mechanical fact about the metric, not advice about what is worth pursuing — a nofollowed mention in a publication your buyers read is still worth more than a followed link in a directory.
4. If you use Semrush, fix the six spam signals too. Authority Score weighs eight factors, and six of them are manipulation checks: no organic rankings, an unnaturally high share of dofollow domains, an imbalance between links and traffic, referring domains sharing an IP address, referring domains sharing an IP network, and another domain with an identical backlink profile. This is the only one of the four scores where cleaning up moves the number.
5. Wait, and expect the score to move without you. Moz refreshes its link index continuously and recalculates the model periodically; Semrush updates Authority Score "once every two weeks"; Ahrefs refreshes its index "every 15–30 minutes." A number that moves on three different clocks is not a weekly KPI.

What does not work, despite being sold constantly: buying links on high-scoring sites, link exchanges, and paying for a "DA boost" package. Ahrefs says the quiet part out loud about why the last one is even possible: "DR doesn't account for backlink SPAM. In fact, large amounts of low-quality backlinks may actually increase your DR, not decrease it." You can absolutely buy a higher number. You are buying the number, not the thing it was built to approximate, and Google's spam policies treat the purchase itself as the violation.

Why domain authority exists at all: the 4,678-day hole#

Every folklore has an origin, and this one has a date.

For roughly a decade Google published an authority number for your website, in public, for free. It was Toolbar PageRank: a 0 to 10 integer, visible in the browser, updated a few times a year. Link builders priced guest posts off it. Agencies put it in pitch decks. It was crude and it was gameable and it was, unambiguously, Google's own number about your whole site.

Google updated it for the last time on 6 December 2013. Then, as Search Engine Roundtable reported on 9 March 2016: "On Monday, March 7 2016, Google officially killed off Toolbar PageRank scores to the few browser tools and web site tools that use it." The same report notes the remaining data "will be completely unavailable to all within two weeks."

Count the days from that last update to today and you get 4,678. Twelve years and nine months in which the only organisation that knows how your site is evaluated has published no site-level number about it whatsoever.

Now count the other side of it, which I have not seen anyone do. The Google Toolbar shipped on 11 December 2000. From that day to the final update is 4,743 days of Google publishing a public authority number for your website. From the final update to today is 4,678 days of silence.

The two spans are 65 days apart. On 1 December 2026, the silence becomes longer than the signal ever was, and an entire professional discipline will have spent more of its life without Google's number than with it.

That is the hole. Markets do not tolerate holes where a number used to be. Four companies filled it, and because each one built its own crawler and its own model, they filled it four different ways. The industry then did the thing industries always do with a substitute that arrives during a shortage: it forgot it was a substitute.

Here is the part that matters for your decisions. A proxy adopted in a vacuum has no correction mechanism. When Toolbar PageRank was wrong you eventually found out, because it came from the scorer. When a third-party authority score is wrong there is nothing to check it against, so it cannot be wrong — it can only be the score. Thirteen years of that is how a predictive model becomes a target.

A figure comparing two spans of time on one shared linear scale, to show that Google's public authority number has now been gone for almost exactly as long as it existed. The upper bar covers the 4,743 days from 11 December 2000, when the Google Toolbar began displaying a public 0 to 10 PageRank score for any website, to 6 December 2013, when that score was updated for the last time. The lower bar covers the 4,678 days of silence since, with no public authority number from Google at all. Because both bars are drawn off the same maximum, they are almost the same length: the difference is 65 days, and the figure sets that number large to make the point that on 1 December 2026 the silence becomes longer than the signal ever was. The conclusion stated on the figure is that the substitute metrics were adopted into a vacuum that is about to be older than the thing it replaced. Day counts calculated by the author from Google's own published dates, counted to 27 September 2026, the day of writing.

Google published a public authority number for your site 4,743 days 11 Dec 2000 → 6 Dec 2013

Google has published nothing 4,678 days 6 Dec 2013 → 27 Sep 2026

Both bars on the same scale. 65 days short of the bar above

65 days until the silence is longer than the signal ever was On 1 December 2026 the silence becomes the longer of the two. Toolbar PageRank last updated 6 Dec 2013; the public score was formally killed on 7 Mar 2016.

Google published a number 4,743 days

Google published nothing 4,678 days

Same scale. The gap is 65 days.

65 days until the silence is longer than the signal was On 1 Dec 2026 the silence becomes the longer of the two.

How to use domain authority in 2026, in five steps#

Nothing is held back for later. This is the whole method, and the rest of the post is the evidence underneath it.

1. Pick one tool and never mix two. Moz's Domain Authority, Ahrefs' Domain Rating, Semrush's Authority Score and Majestic's Trust Flow are computed from four different crawls with four different models. A domain reading 58 in one and 41 in another has not changed; you have changed instrument. Mixing them inside a single prospect list silently sorts your list by which tool happened to crawl which site. The four scores are broken down below.
2. Use it as a sort, never as a gate. Rank your prospect list by the score, then ignore the number and open the pages. The metric's only defensible job is deciding what you look at first. The moment it decides what you contact, you have handed a third-party model a veto over your business development. Both vendors that sell one of these scores say in writing that no absolute value is good or bad.
3. Delete the threshold from your brief. "DA 40+" is the single most copied instruction on this keyword and it comes from the three pages with nothing to sell, not from the vendors. Replace it with a relative test: is this site comparable to, or better than, the sites already linking to the page I am trying to move?
4. Check the score's inputs before you trust it on a specific site. A score built on referring domains alone will rate a site with 4,000 spam links above a trade publication with 90 real ones. A score that mixes in traffic will do the reverse. Know which one you are holding, because the failure modes are opposite and both look like a number.
5. Score the page, not just the domain. The link you are chasing lives on a page, and what predicts whether it passes anything useful is whether that page is read, indexed and linked to by anyone. Domain-level scores cannot see that. A weak domain's genuinely-read resource page is worth more than a strong domain's orphaned link dump, and Google's own documentation is closer to this view than to yours.

That is the working method. Everything below is why each step is there.

What Google has actually said about a website authority score#

The cleanest statement is also one of the oldest. In a Webmaster hangout in October 2016, at the 42:47 mark, John Mueller said eight words that the entire consensus rests on, reported at the time: "we don't have anything like a website authority score."

Take that at face value. It is unambiguous, it is on the record, and Google has never retracted it.

Now put it next to the strongest counter-claim on this SERP, which comes from the vendor page sitting at position four. Ahrefs' website authority checker says: "Google representatives consistently deny the use of 'domain authority' as a ranking factor. But according to Google's John Mueller, they do have a sitewide score that 'maps to similar things.'" Ahrefs links to no source for it.

I went and found the source. It is a SearchLove interview from late 2020, and what Mueller actually said is this:

I don't know if I'd call it authority like that, but we do have some metrics that are more on a site level, some metrics that are more on a page level, and some of those site-wide level metrics might kind of map into similar things.

Read the two versions next to each other, because the difference is the entire disagreement in miniature.

Mueller said I don't know if I'd call it authority, some metrics, might, kind of, similar things. Five hedges in one sentence, from a man being careful. Ahrefs renders that as "they do have a sitewide score that 'maps to similar things'" — the hedges deleted, the indefinite made definite, a metric promoted to a score, and the whole thing set in quotation marks.

Nobody lied. A careful sentence was simply passed through a company that sells the thing it is about, and came out harder than it went in. That is how the folklore on this keyword is manufactured: not by invention, by compression. It is worth knowing that the same mechanism is running on every other quote you will read about this.

Here is what is not unsourced, and it is stronger than the quote anyway.

Google's own published spam policies contain a section called Site reputation policy. Its opening line defines the abuse as third-party content published on a host site "mainly because of that host's already-established ranking signals, which it has earned primarily from its first-party content."

Every load-bearing word there is about the site, not the page. Host's. Already-established. Earned. Google is not being coy: it is describing a property that accrues to a domain over time, that transfers to content published on that domain, and that is valuable enough to rent. That is the thing people are gesturing at when they say domain authority. Google simply does not publish it, does not score it 0 to 100, and does not call it that.

It is not only the policy page, either. Among the internal Google documents released as exhibits in the US antitrust trial is a deck headed "Search Quality Has Many Aspects", and the list underneath it runs: "Relevance · Page quality · Popularity · Freshness · Localization · Language · Centrality · Topical diversity · Personalization · Web ecosystem · Mobile friendly · Social fairness · Optionalization · Porn demotion · Spam · Authority · Privacy · User control of spell correction", followed by the line "Capturing everything in a metric is tough!" (exhibit UPX0204, "Ranking for Research"). Authority is on Google's own internal list. Note also what sits two items above it: page quality, not site quality.

So the honest position is narrower than either side wants. Google has no metric named domain authority and does not read Moz's. Google unmistakably has site-level signals and an internal notion of authority, by its own description, in its own documents. Neither of those facts tells you the number in your dashboard is close to them.

What the leaked Google documentation calls it: siteAuthority#

In March 2024 a large set of internal Google Search API documentation was published to GitHub, apparently by accident, and removed on 7 May 2024. Rand Fishkin wrote it up on 28 May as "more than 2,500 pages of API documentation containing 14,014 attributes (API features)." Mike King's technical breakdown counts "2,596 modules represented in the API documentation with 14,014 attributes (features)."

You do not have to take either writeup's word for any of it, which is the part that gets lost. The leaked material was Google's own Elixir client library, published by Google to a public package registry, and the generated reference is still browsable today. Here is the relevant line from the CompressedQualitySignals module, verbatim:

siteAuthority ( type: integer() , default: nil ) - site_authority: converted from quality_nsr.SiteAuthority, applied in Qstar.

The module's own description, on the same page, says it carries "per doc signals that are compressed and included in Mustang and TeraGoogle." Its neighbours in that module are named authorityPromotion, unauthoritativeScore and experimentalQstarSiteSignal. Elsewhere, in PerDocData, hostNsr is described as "Site rank computed for host-level sitechunks."

An integer. Named site authority. Converted from a site-level quality signal. Applied in a ranking system.

Google did not say the documents were fake. It confirmed them. The Verge reported on 29 May 2024: "A collection of 2,500 leaked internal documents from Google filled with details about data the company collects is authentic, the company confirmed today." The only substantive comment, from spokesperson Davis Thompson, was this:

We would caution against making inaccurate assumptions about Search based on out-of-context, outdated, or incomplete information. We've shared extensive information about how Search works and the types of factors that our systems weigh, while also working to protect the integrity of our results from manipulation.

Now the discipline, because this is where the SEO industry lost its head in 2024. A named attribute in an API surface is not proof of a live ranking factor. Documentation describes what can be stored and passed around, not what is weighted, not how much, and not whether the field is still populated. "Outdated" and "incomplete" in that statement are doing real work and deserve to be taken seriously rather than sneered at.

What the attribute does establish is narrow and still decisive: a Google-published document, whose authenticity Google itself confirmed, contains a site-level integer called site authority. That is enough to retire the flat claim that no such concept exists anywhere at Google — which is the claim nine of the ten pages on this keyword are built on, and which not one of those ten pages mentions the leak in connection with.

It is not remotely enough to make your Moz score meaningful. siteAuthority is an internal value on Google's own crawl of the entire web. Your dashboard holds a vendor's model of a vendor's crawl. Nothing has ever been published that measures the distance between the two, because only one side can see both.

What the ten pages ranking for "domain authority" actually say#

On 27 September 2026 I pulled the ten pages currently ranking for this keyword, stripped out the navigation, the footers and the signup forms, and read the 20,578 words that were left, classifying each page on nine questions. Wikipedia, Network Solutions, Wix, Ahrefs, Crazy Egg, MonsterInsights, Semrush, Digital Strike, G2 and Surfer.

The headline result is agreement, which is rare enough to be worth stating. Nine of the ten tell you plainly that domain authority is not a Google ranking factor. The consensus on this keyword is correct. Whatever else is wrong with the page you land on, it is not going to tell you Google reads your Moz score.

A figure reporting a hand count of the ten pages ranking for the query domain authority, which finds that the consensus on this keyword is precise about what the metric is not and silent about what it is. Nine of the ten pages state that domain authority is not a Google ranking factor or not a Google metric. None of the ten mentions the May 2024 Google API leak or the siteAuthority attribute, and none of the ten gives any correlation coefficient between an authority score and actual rankings, so both of those rows are drawn as completely empty. Three of the ten name a specific numeric target score to aim for, and two of the ten warn that the metric can be gamed or inflated. Each question is drawn as a row of ten unit cells, filled where a page met the criterion and hollow where it did not. The ten ranking pages were pulled on 27 September 2026, stripped of navigation, footers and forms, and the 20,578 remaining body words were classified by hand.

States it is not a Google ranking factor 9 of 10

Mentions the May 2024 Google API leak 0 of 10

Gives a correlation coefficient vs rankings 0 of 10

Names a specific numeric target score 3 of 10

Warns the metric can be gamed or inflated 2 of 10

Nine pages agree on what domain authority is not. None cites a coefficient for what it is. The ten pages ranking for the query were pulled on 27 September 2026, stripped of navigation, footers and forms, and the 20,578 remaining body words were classified by hand.

Not a Google ranking factor 9 of 10

Mentions the May 2024 API leak 0 of 10

Gives a correlation coefficient 0 of 10

Names a numeric target score 3 of 10

Warns it can be gamed 2 of 10

Nine agree on what it is not. None quantifies what it is.

Two other counts are more interesting.

Zero of the ten give a correlation coefficient between any authority score and actual rankings. Every one of those pages is built on the premise that a higher score means better ranking odds, and across 20,578 words the strength of that relationship is quantified nowhere. Two pages offer statistics of any kind. Ahrefs cites a study of 218,713 domains and says the two "correlate well" without publishing a number, and adds the caveat itself. Wikipedia gives ρ ≈ 0.9, but read what it is measuring: that is how closely the three vendors agree with each other, not how any of them tracks Google.

That distinction is the whole ballgame and almost nobody draws it. Three tools built on three separate crawls converging on nearly the same score tells you the tools share a method. It tells you nothing at all about whether the method predicts anything.

The coefficient exists. Moz published it. It is 0.12.

Here is the part that turns a gap into a story.

The number those ten pages do not print has been sitting in public since 5 March 2019, on Moz's own blog, published by Moz. Russ Jones ran the validation for Domain Authority 2.0 and wrote the method out: 16,000 randomly selected US keywords, the full first page collected for each, then "the mean Spearman correlation coefficient for each SERP" calculated for Moz's DA, Majestic's Citation Flow and Trust Flow, and Ahrefs' Domain Rank.

The result, in Moz's words:

Moz's Domain Authority scored a ~.12, or roughly 6% stronger than the next best competitor (Domain Rank by Ahrefs.) Domain Authority performed 35% better than CitationFlow and 18% better than TrustFlow.

0.12. And that is the winner. Moz's summary of the same study reads: "Moz's new Domain Authority has the strongest correlations with SERPs of the competing strength-of-domain link-based metrics in the industry."

Both sentences are true. The strongest domain-strength metric in the industry has a mean per-SERP Spearman correlation with rankings of about 0.12, which means that inside a single set of ten results the score barely orders them at all. Moz is candid about it in the same post, volunteering that "at first blush, Domain Authority's improvements over the competition are, frankly, underwhelming," and stating the causation boundary itself: "Domain Authority does not cause search rankings. It is not a ranking factor. Domain Authority predicts the likelihood that one domain will outrank another."

Two caveats I will not skip. The study is seven years old and describes DA 2.0 at launch; Moz has not published an updated figure, and the current Domain Authority page carries no coefficient at all. And a low within-SERP correlation is partly an artefact of the question: by the time ten pages are competing for one keyword, Google has already filtered on relevance, so the remaining variance is small by construction.

Neither caveat rescues the threshold. The only coefficient any vendor has ever published for this class of metric is 0.12, it was published by the vendor with the most to gain from a bigger one, and not one of the ten pages ranking for the keyword mentions it.

Zero of the ten mention the leaked Google documentation, or the attribute in it called siteAuthority. More on that below. Network Solutions gets closest, and goes the other way: "Google has repeatedly confirmed that it doesn't use any kind of 'site authority' score in its ranking algorithm." That is the exact assertion the leak complicated, stated on the SERP as settled.

The inversion nobody mentions

Now the finding I did not expect, and the one you can act on immediately.

Three of the ten pages hand you a number to aim for. None of those three sells an authority metric. Both of the pages that do sell one refuse to give you a number.

The three with targets are a domain registrar, a website builder and a WordPress analytics plugin, and all three land on roughly the same band: 40 to 50 average, 50 to 60 good, above 60 excellent. None of them attributes that band to anything.

The two vendors say the opposite in writing. Ahrefs, which sells Domain Rating: "It's not possible to say that a Domain Rating of 30 is good, or 50, or 60, or 70." Semrush, which sells Authority Score: "There's no single number that counts as 'good' for every site."

A figure splitting the same hand count by who published the page, which finds that the companies selling an authority metric are the only ones that refuse to name a good score. Of the two pages published by a vendor that sells such a metric, Ahrefs and Semrush, zero name a target score to aim for, and both refuse in writing: Ahrefs says it is not possible to say that a Domain Rating of 30 is good, or 50, or 60, or 70, and Semrush says there is no single number that counts as good for every site. Of the eight pages published by someone with no authority metric to sell, three do name a target, and those three are a domain registrar, a website builder and a WordPress analytics plugin, all converging on the same unattributed band of 40 to 50 average, 50 to 60 good and 60 or above excellent. The remaining five non-vendor pages name no target. Each group is drawn as a row of unit cells, filled where the page names a number. From the hand count of the ten pages ranking for this keyword on 27 September 2026.

Published by a vendor that sells an authority metric 0 of 2 name a target score

Ahrefs “It’s not possible to say that a Domain Rating of 30 is good, or 50, or 60, or 70.” Semrush “There’s no single number that counts as ‘good’ for every site.”

Published by someone with no metric to sell 3 of 8 name a target score

The number everyone repeats 40–50 average. 50–60 good. 60+ excellent. Unattributed.

The three: a domain registrar, a website builder and a WordPress analytics plugin. The other five non-vendor pages name no target either.

A target number for a deliberately relative metric reverses the vendors’ own position.

Sells an authority metric 0 of 2

Ahrefs “It’s not possible to say that a Domain Rating of 30 is good, or 50, or 60, or 70.”

Semrush “There’s no single number that counts as ‘good’ for every site.”

No metric to sell 3 of 8

The three: a domain registrar, a website builder and a WordPress analytics plugin.

All three name one band 40–50 average, 50–60 good, 60+ excellent. Unattributed.

The sellers name no number. The non-sellers name one.

The companies with the most commercial incentive to make you chase a score are the ones telling you there is no score to chase. The companies with no stake in the metric are the ones printing the ladder. A target number for a relative metric is not a simplification of the vendors' position, it is a reversal of it — and it is the single most copied sentence on this keyword.

If you are reading this because a brief on your desk says DA 40+, the honest fix is not a better threshold, it is an outreach process that judges sites by reading them instead of by sorting them.

The four domain authority scores, and what each one actually counts#

There is no such thing as the domain authority score. There are four, they are computed from four separate crawls of the web, and they are not measuring the same thing. Which one you hold changes what your whole link building programme optimises toward, so it is worth thirty seconds to know. Anyone quoting a threshold without naming the tool is quoting a number with no units on it.

Moz, Domain Authority. Moz's own page, last updated 18 March 2025, puts the scale at "1 to 100" and the method as "a machine learning algorithm that evaluates multiple factors, primarily focusing on backlink data." It is explicitly comparative: "It's relative: DA compares your website against others, so it's most useful for benchmarking your performance within your industry."

Ahrefs, Domain Rating. Ahrefs defines DR as "the strength of the website's backlink profile on a logarithmic scale from 0 to 100." Links, and only links: the metric "doesn't take into account any other variables like link spam, traffic, domain age, etc." Ahrefs is also the only one that spells out what logarithmic does to your plans: "the gap between DR 75 and DR 76 is much bigger than between DR 20 and DR 21."

Semrush, Authority Score. Semrush's differentiator is that it deliberately looks past the link graph, and it markets that fact against the others: "A site with a large but manipulated backlink profile … will score lower with Authority Score than it would with a backlink-only metric. That makes it harder to game." Its free checker also states the boundary plainly: "Website authority is not an official metric used by Google, AI search platforms, or other search engines."

Majestic, Trust Flow and Citation Flow. Scale 0 to 100, both. Majestic refuses to produce one number at all. Its glossary defines Trust Flow as "a score based on quality, on a scale between 0-100" built from "many trusted seed sites based on a manual review of the web," so that "sites closely linked to a trusted seed site can see higher scores." Citation Flow is "a score between 0-100 which helps to measure the link equity or 'power' the website or link carries," and it "predicts how influential a URL might be based on how many sites link to it." Quality and quantity are reported as two separate axes precisely because collapsing them loses the information you needed.

Line those up and the disagreement is structural, not noise. One vendor models rankings from links. One measures link strength and nothing else, and says outright that spam can raise it. One weighs eight named factors, six of which exist to catch manipulation. One declines to merge quality with quantity at all. Four different answers to four different questions, presented to you as one number in the same column of the same spreadsheet.

Even the floors disagree. Moz runs 1 to 100 and says "new websites always start with a Domain Authority of 1." Semrush runs 1 to 100. Ahrefs and Majestic run 0 to 100. A site reading 1 in one tool and 0 in another is at the bottom of both, but any rule written as "above 30" is a different rule in each of them.

Two properties of the scale that break every threshold

The first is logarithmic compression, which Ahrefs states and Moz describes in different words: "it's easier to grow your score from 20 to 30 than it is to grow it from 70 to 80." So a brief that says "get us to DA 50" is not a target, it is a target whose cost depends entirely on where you start, by an amount nobody has quantified for you.

The second is worse, and Moz volunteers it. The scale is renormalised against the whole index. In Moz's own example: "If Facebook.com acquired a billion new links, other site's DA would drop relative to Facebook's." Because "more established and authoritative domains like Facebook will have extensive link profiles, they take up more high-DA slots, leaving less room at the higher end of the scale."

Read that as an operator. Your score can fall while your site improves, because someone else's site improved more. It is a percentile in disguise. Reporting a percentile as a KPI, to a client, without that caveat, is how this metric ends up in contracts it cannot survive.

What domain authority is honestly good for#

None of the above means you should close the tab on the metric. It means you should stop asking it a question it cannot answer and start asking it the one it can.

The question it cannot answer: will a link from this site move my rankings? The only coefficient any vendor has ever published for this class of metric is Moz's own ~0.12, and Moz states in the same post that the score "does not cause search rankings." Ahrefs' own blog puts the boundary in one line: "Google ranks pages, not websites."

The question it can answer: of these 400 sites, which 40 should I open first? That is triage, it is the same job a score does in competitor backlink analysis, and the scores are genuinely good at it. A site sitting at DR 6 with eleven referring domains and a site at DR 71 are different propositions, and you do not need a coefficient to act on that. You need an ordering, and an ordering is exactly what a relative metric produces.

The distinction is not academic. It changes what the number is allowed to do in your process:

• As a sort key, the score costs you nothing when it is wrong. You open the pages in a slightly suboptimal order and your eyes correct it within seconds.
• As a gate, the score's errors become invisible and permanent. Every site it scored badly for a reason that does not apply to you — a young domain with real editorial reach, a niche trade publication with ninety links that all matter, a regional outlet that outranks you for everything you care about — never makes it to a human. You will never see the ones it removed. That is what makes a gate worse than a bad sort rather than a stricter one.

And one practical note that costs nothing. If you must report a score to someone, report it next to the same score for three named competitors, from one named tool, on a stated date. In that form the number carries its own units and its own caveat, and it stops being a promise.

The limits of this evidence#

Four things you should weigh against everything above.

The hand count is mine and it is a snapshot. Ten pages, 20,578 words, counted on 27 September 2026, from one search, from one location. Results differ by country, by personalisation and by week. I have named all ten publishers in the text so you can repeat it. Where a page was genuinely ambiguous on a question I counted it as a no, which pushes the counts down, not up.

The 4,743 and 4,678 day figures are my arithmetic, derived from three published dates: the Toolbar's 11 December 2000 release, its final 6 December 2013 update, and today. The dates are sourced; the subtraction is mine.

The leaked documentation is not a ranking factor list. I have argued that above and I will not soften it: an attribute in an API surface tells you a field exists, not that it is weighted, not how much, and not that it is still live. Google's own response asked readers not to draw conclusions from "out-of-context, outdated, or incomplete information," and that request is reasonable.

And the strongest counter to my own reading of that 0.12 is a statistical one. Within a single SERP, Google has already filtered on relevance, which compresses the variance a domain metric could possibly explain. A metric can be genuinely informative across the whole web and still barely order ten finalists. If you think that rescues the threshold, I would say it does the opposite: it means the score is at its least useful precisely where people apply it hardest.

Several of the Google statements here are secondary. Mueller's and Illyes' remarks were made on Twitter, on Reddit and in interviews, and I am relying on contemporaneous SEO reporting for the wording and the dates because the originals are behind login walls or inside unindexed video. Where I could reach a primary document — the spam policy, the trial exhibit, the API reference, The Verge's confirmation, Moz's own study — I quoted the primary document instead, and you can check every one of them from the links above.

The shape of the whole thing#

Google published an authority number for your site for 4,743 days and has published nothing for 4,678. Four companies filled the gap with four different models of four different crawls, and every one of them says in writing that its number is only meaningful in comparison. Google says it has nothing like a website authority score; Google's own spam policy describes a host site earning ranking signals; an internal Google deck lists Authority among the aspects of search quality; and Google-published documentation, which Google confirmed is real, contains an integer called siteAuthority. All of those statements can be true together.

What cannot be true is the sentence the industry actually runs on: that there is a number, that 50 is good, and that getting it up is a project. Nine of the ten pages ranking for this keyword will tell you the score is not Google's. Three of them will then hand you a threshold, and none of the three sells the thing they are giving you a threshold for. None of the ten mentions that the only published coefficient is 0.12.

Use it to decide what to read first. Never to decide what to pursue. The sites that will actually link to you are not distributed by score, they are distributed by whether anyone at that organisation has a reason to care about the page you wrote. No metric on the market can see that, and finding out costs an email.

Frequently asked questions#

What is a good domain authority score? There isn't one, and the two companies that sell an authority score both say so in writing. The useful reframe: a score is good if it is comparable to the sites already linking to the page you want to outrank. Export the referring domains of the current top result, look at the spread of their scores, and that spread is your answer for that keyword. It will be different for a dentist in Leeds and for a developer tools startup, which is exactly why no universal band exists.

Is domain authority the same as domain rating? No, and mixing them is the most common practical error on this topic. Domain Authority is Moz's, built with a machine learning model on Moz's index. Domain Rating is Ahrefs', a logarithmic measure of backlink profile strength on Ahrefs' index that deliberately excludes traffic, spam signals and domain age. They will disagree on the same site, often by twenty points or more, and neither is wrong, because they are answering different questions on different data.

Why did my domain authority drop when nothing changed on my site? Most often because the scale moved, not your site. These scores are normalised across the whole index, so a competitor's growth, a recrawl, or an index refresh can move your number while your backlink profile is untouched. Before investigating a drop, check three things in order: whether your referring domain count actually fell, whether the tool shipped a model update that quarter, and whether the drop shows up in the other three tools. If only one tool moved, the tool moved.

How long does it take to increase domain authority? Longer at the top than the bottom, by design. Because the scale is logarithmic, the same effort that lifts a new site from 10 to 20 will barely register at 60. A more honest planning frame is to stop forecasting the score at all and forecast the input instead: how many editorially-given referring domains can you realistically earn per quarter, and from what kind of site? That number you can plan against, and it is the number the score is a lagging approximation of.

Does domain authority matter for AI search and LLM citations? No vendor has published evidence that it does, and the mechanics differ enough that you should assume nothing carries over. Assistants cite pages that directly answer a question and that they can retrieve and parse, which is closer to a relevance-and-clarity problem than a site-strength one. Semrush's own checker states plainly that website authority "is not an official metric used by Google, AI search platforms, or other search engines." If someone sells you an authority score as an AI visibility lever, ask for the study.

Do I need to pay for a tool to check domain authority? No. Moz, Ahrefs and Semrush all run free authority checkers, rate-limited but adequate for spot checks, and Majestic publishes Trust Flow and Citation Flow. The catch is not price, it is consistency: pick one, and record the tool name and the date alongside every score you write down. A score without a tool and a date attached is not data, and six months later you will not remember which column it came from.

Should I buy a link from a high domain authority site? Setting aside that paid links without qualification are a direct violation of Google's spam policies and a fast route to the kind of profile people later panic about as toxic backlinks, the score is a bad reason to do it even on its own terms. A high site-level number tells you nothing about whether the page your link sits on is indexed, read, or linked to by anyone. Plenty of high-authority domains host link inventory nobody has visited in years. The check that actually matters is whether that specific page has an audience, and no domain-level metric can see it.