Page-Level Authority Metrics and When They Beat Domain Scores

Page-level authority metrics — Ahrefs’ URL Rating, Moz’s Page Authority, and similar scores elsewhere — estimate the link strength of one URL rather than a whole domain. They exist because search results are pages, and a domain score averages over everything a domain has ever published.

For most questions people ask link data, the page-level number is the more relevant one and the domain-level number is the one they quote.

What the two scores are describing

A domain score summarises inbound links to the domain as a whole. It’s a property of the site’s aggregate link position. It says nothing about which pages hold that position.

A page score summarises links pointing to one URL — usually external links plus, in some implementations, internal links, which is a meaningful design difference. Vendors document this at a high level; the weights are closed either way.

The practical difference: a large publisher may have a very high domain score and an individual article with almost no links to it. Both facts are true simultaneously, and which one is relevant depends entirely on your question.

The three questions where page-level wins

“Why is this page not ranking?” You are competing with URLs. If the pages outranking yours have substantially more inbound links than yours does, that’s a page-level observation, and the domain-level comparison may point the other way entirely. This is the correct unit for the exercise in how many backlinks do you need to rank.

“Is this a good page to be cited from?” A link from a well-linked article on a mid-sized site and a link from an orphaned page on a large site are different things. Domain score cannot distinguish them; page score attempts to.

“Where is our link equity concentrated?” Sorting your own pages by page-level score shows which pages the outside world actually cites. That is frequently a surprise — often the most-cited page is an old post nobody in the company thinks about.

The two questions where domain-level wins

Bulk triage. Sorting a thousand prospect domains needs one number per domain, and page-level scores for pages that don’t exist yet are undefined. Domain score is the right tool here, with all the caveats about it being ordinal and index-dependent.

Comparing organisations. “How does our site’s link position compare to these four competitors” is a domain-level question by construction.

Notice that both are comparative and coarse. That’s the shape domain scores fit.

Why page scores are noisier

Three reasons, all worth knowing before you put one in a slide.

Smaller n. A domain score aggregates thousands of links; a page score may rest on eleven. Small sets move sharply on single events — the general problem in small link sets and statistical limits.

More index sensitivity. One new link discovered, or one stale link retired, is a large relative change to a page with few links. A domain absorbs the same event invisibly.

URL identity problems. A page score is attached to a URL string. Trailing slashes, protocol variants, www, tracking parameters, and case differences can split one page’s links across several URL rows depending on how the vendor normalises. Check whether your page’s score is fragmented before concluding it’s low.

That last one is the most common false alarm. A page with a genuinely healthy link position can show a poor page-level score simply because half its links point at a variant the vendor tracks separately.

Reading both together

The interesting information is usually in the relationship between the two scores, not either alone.

• High domain, low page. Common for new or deep pages. Tells you the page hasn’t been cited yet; the site’s overall position may still help it. This is the ordinary state of most pages on most sites.
• Low domain, high page. Something specific was cited — a tool, a dataset, a viral post. Worth understanding, because whatever caused it is repeatable in a way most link acquisition isn’t.
• Both low, in a competitive set where rivals are both high. A genuine link-position gap.
• Both high, still not ranking. Links are probably not the constraint. This is one of the most valuable findings link data produces, and you can only see it by looking at both levels.

A hypothetical table for one target query: your page at page-score 12 on a domain-score 41 site, against competitors at page-scores 31, 28, 44 on domain scores 38, 52, 49. The page-level column tells a consistent story the domain-level column doesn’t. (Illustrative figures, not measurements.)

How to report it

Name the metric, vendor, and level explicitly. “Ahrefs URL Rating (page-level), pulled 17 July” — not “authority.”

Never mix levels in one comparison. Comparing your page score to a competitor’s domain score is meaningless, and it happens often because the domain number is easier to find.

Give the referring-domain count next to the score. A page score is a summary; the count of distinct linking domains to that URL is closer to a raw observation, and the pair is more informative than either.

Check URL normalisation before reporting a low score. One line in the report: “score reflects the canonical URL; variant URLs checked and hold no material links.”

Don’t do arithmetic across the scale. Page scores are compressed at the top for the same reason domain scores are; percentage changes and averages have the same problems described in Domain Rating versus Domain Authority.

What nobody outside the vendors knows

Whether internal links are included and how they’re weighted. How the score handles a page that redirects. How quickly it responds to a newly discovered link. Whether the compression curve is the same as the domain-level one. And, as always, how any of it relates to whatever page-level signals a search engine computes — which are not published and are not these.

The reliable part is the framing: search results are pages, so page-level link data is usually the right unit, and the habit of checking it is what stops a report from confidently describing a domain when the question was about a URL.