Does domain authority actually matter?
Google does not use it. It still predicts something real. Both facts are true, and confusing them is why so much SEO budget goes to the wrong places.
Two statements, both true, routinely presented as if only one can be.
Google does not use domain authority. It is not a ranking factor, there is no equivalent internal metric, and Google representatives have said so for years.
Domain authority also predicts something real. Sites with higher scores do tend to rank better, the correlation is not an illusion, and dismissing the metric entirely leads people to ignore a constraint that genuinely governs what they can win.
Reconciling these is not a debating point. It determines whether you spend a year on queries you cannot reach.
Key takeaways
- It is a third-party estimate, built by modelling ranking outcomes — a thermometer, not the weather.
- The correlation is real because links are real. The score is a proxy for links from sites with standing.
- Absolute values are meaningless. Only the comparison against the sites you are competing with matters.
- The scale is logarithmic. 20 to 30 is ordinary work; 70 to 80 is a different category of undertaking.
- Optimising the score directly is the trap — it is an output, and moving it without moving the cause achieves nothing.
What the score actually is
Domain authority, domain rating and their equivalents are all the same kind of object: a tool vendor’s model that takes a site’s backlink profile and produces a number, typically 0 to 100, intended to predict how well that site can rank.
They are built by training against observed rankings. The vendor gathers link data, gathers ranking data, and fits a model that outputs a score correlating with where sites actually appear. This is a reasonable thing to build, and it is why the correlation exists — the score was constructed to correlate.
Which also explains why it is not a ranking factor. Google is not reading the vendor’s number. The vendor is reading Google’s results and reverse-engineering a number that approximates them. Cause runs the opposite way from how it is usually discussed.
The useful analogy: a thermometer correlates strongly with how hot it is, and breaking it does not cool the room. Domain authority correlates strongly with ranking ability, and raising it artificially does not create the ability.
Why Google does not use it
Beyond the obvious point that Google has no reason to consume a competitor’s metric, there is a substantive reason no site-wide authority score would work well.
Authority is not a property of a domain. It is a property of a domain in relation to a subject. A national newspaper has enormous general standing and very little standing on the specifics of industrial adhesives. A three-year-old site run by two people who have written the definitive resources on industrial adhesives has the reverse.
A single number per domain cannot represent that, which is why Google’s systems evaluate relevance and quality at the level of pages and topics rather than assigning sites a global rating. It is also why the score sometimes misleads badly in exactly the situations where you most need guidance: specialist queries in narrow fields, where a low-scored specialist regularly outranks a high-scored generalist.
What it is genuinely useful for
Despite all that, throwing the metric away is a mistake. It has two real uses.
1. Deciding whether a query is worth attempting. This is the important one. Search your target query, look at the scores of the ten results, and compare with yours. If they sit at 75 and you are at 18, that query is not a this-year project no matter how good your page is. If they sit at 30, it is entirely available.
This single check prevents the most expensive strategic mistake in SEO — committing a quarter to terms the site cannot reach. It takes two minutes.
2. Comparing link opportunities. When choosing where to spend outreach effort, the score is a crude but fast filter. Crude because relevance matters more, but useful for excluding sites with no standing at all.
Both uses are relative. Neither involves caring about your own number in isolation, which is the form the question almost always takes and the form in which it is least useful.
Four ways it misleads
It ignores topical relevance entirely. A link from a general news site scoring 80 may be worth less for a specialist query than one from a trade publication scoring 35 that your entire industry reads. The score has no way to express this, and it is the difference that most often decides specialist rankings.
The scale is logarithmic, and people read it as linear. Moving from 20 to 30 is a few months of ordinary work. Moving from 70 to 80 is a multi-year institutional undertaking. Treating a ten-point gap as a fixed quantity produces wildly wrong estimates in both directions.
It is a site-wide number applied to page-level competition. Individual pages on a strong domain vary enormously in their own standing. A new page on a high-authority site does not inherit the full benefit, which is why large publishers still fail to rank for terms their score suggests they should own.
It is manipulable, which makes it unreliable exactly where it matters. Because it derives from links, it can be inflated by acquiring links that raise the number without raising real standing. This is an entire industry. It means a high score on an unfamiliar site is weak evidence, particularly when evaluating a link opportunity someone is selling you.
How the number actually moves
Understanding the mechanics removes most of the anxiety attached to the score, and explains several behaviours that otherwise look like errors.
It moves in steps, not smoothly. The score updates when the vendor recrawls and recalculates, which happens periodically rather than continuously. A link earned today may not register for weeks. This produces the common experience of doing nothing for a month and seeing the number jump, then doing a great deal and seeing it sit still — neither reflects what happened in that period.
It can fall without you doing anything wrong. Because the scale is relative and recalibrated against the wider web, a static site drifts downward over time as the rest of the web accumulates links. A score that fell two points while you published steadily usually means the index grew, not that you lost standing.
New links have front-loaded effect. Your first genuine link from a relevant site with real standing moves the number noticeably. The fiftieth from a similar site moves it barely at all. This is the logarithmic scale again, and it is genuinely good news for new sites — early progress is fast and visible, which is precisely when encouragement is most useful.
Losing links is normal. Pages get deleted, sites go offline, articles are rewritten. A referring domain count that falls slightly month to month is the ordinary decay of the web rather than a problem to solve. Chasing every lost link is a poor use of time compared with earning new ones.
The practical upshot: check it quarterly, not weekly. It is a slow-moving relative estimate updated in batches, and reading it more often than it meaningfully changes generates false signals that lead to real decisions. If you want a weekly number, use impressions from Search Console — that one actually reflects what happened last week.
What to look at instead
For the decisions the score gets used for, better inputs exist.
| Question | Better signal than DA |
|---|---|
| Can I rank for this query? | Referring domains of the ranking pages, and how well they match intent |
| Is this link worth pursuing? | Does the site have real readers in my field? Would I want it without the link? |
| Am I making progress? | Referring domains over time; count of queries in the top ten |
| Which page should I improve? | Impressions and position from Search Console, not site-level metrics |
| Is my site stronger than last quarter? | Non-branded organic clicks, which is the outcome the score is a proxy for |
Notice that most rows replace a site-wide estimate with a page-level observation. That is the general correction: authority questions feel site-wide, but ranking happens page by page, and the closer your measurement sits to the page the more it tells you. Our guide to analysing competitors covers the page-level version of this comparison.
Whether AI assistants care about it
Considerably less, and this is one of the more consequential differences between the two channels.
When an assistant answers a question, it runs searches, retrieves candidate pages, and composes from passages that address the question. The selection at that final stage is driven by whether a specific passage answers the specific question, how current it is, and how cleanly it can be extracted. Site-level authority influences whether the page surfaced in the retrieval step at all, but it does not dominate the choice among retrieved candidates the way it dominates a competitive ranking.
The practical consequence is a real asymmetry for smaller sites. A precise, well-structured, current answer on a domain scoring 20 is frequently cited over a hedged, meandering treatment on a domain scoring 80 — something that essentially never happens in a competitive Google ranking. For a site whose authority gap makes the head terms unreachable this year, that is the channel where the gap matters least. The selection mechanics are covered in how LLMs choose their sources.
So: does domain authority matter? As a diagnostic for what you can realistically win, yes, and ignoring it causes expensive misallocation. As a goal, no — it is an output, and every method of raising it that does not involve genuinely earning standing moves the thermometer without warming the room.
The version worth internalising: use it once, at the start, to decide which queries are worth attempting this year. Then stop looking at it and work on the pages. The score will follow whatever you do to the underlying reality, and watching it in the meantime tells you nothing you could act on.
Find out which queries are actually within reach
Klepha compares you against the sites ranking for your target queries and shows where the gap is authority rather than relevance — plus whether AI assistants already name you.
Run a free visibility reportFrequently asked questions
Is domain authority a Google ranking factor?
No. Domain authority is a proprietary score invented by a third-party SEO tool to estimate ranking strength. Google has stated repeatedly that it does not use it and does not have an equivalent site-wide authority score of its own. The confusion arises because the score is built by modelling actual ranking outcomes, so it correlates with results without being a cause of them.
What is a good domain authority score?
The question has no useful answer in isolation, because the scale is logarithmic and relative. What matters is your score compared with the sites currently ranking for the queries you want. Being at 25 is irrelevant if the results you are competing against sit at 30; it is decisive if they sit at 80.
Can I rank without high domain authority?
Routinely, for queries where competition is low or where existing results poorly match the intent. Authority becomes the deciding factor mainly on competitive commercial terms where every result is already a good answer, so the ranking system needs another basis to choose between them.
How do I increase domain authority?
You cannot increase it directly, because it is an output rather than an input. It rises when you earn links from other sites that themselves have standing. Anything that targets the score rather than the underlying reality — bought links, low-quality directories — moves the number without moving rankings, which is the clearest evidence that the number is not the thing that matters.
Does domain authority affect whether AI assistants cite my site?
Much less than it affects conventional rankings. Assistants select passages that answer the specific question and weight freshness heavily, so a precise answer on a low-authority domain is frequently cited over a vague one on an established site. Authority still helps indirectly, since it affects whether the page ranks well enough to be retrieved in the first place.