The usual complaint about Domain Rating is that it is one signal among many and should not be used alone. That is true and it is also too polite. On the prospect list we opened by hand, DR did not merely fail to help. It ordered the list exactly backwards, and anyone applying a DR floor to it would have kept the useless rows and thrown away the useful ones.
This is not an argument against Ahrefs. DR measures what it says it measures. The problem is the question people ask it.
What we measured
One purchased prospect list, 1,001 rows, 999 distinct URLs across 467 hosts. Every row opened and verified by hand on 7 September 2026. The full study is published with its method and its limits.
| Slice of the list | Rows | Could carry a followed comment link |
|---|---|---|
| Ahrefs DR 50 or above | 285 | 0 |
| Ahrefs DR 10 or below | 460 | 34 of the 37 |
| Whole list | 1,001 | 37, or 3.7% |
The median DR of the rows that worked was 1. The strongest was 29. A DR 30 floor, which is a conservative and entirely standard rule, would have removed every usable row on this list.
Why it inverted: hosted subdomains inherit their platform’s score
The explanation is visible the moment you group the rows by host rather than looking at them one at a time.
| Host pattern | Rows | Ahrefs DR on every one | What was actually on the page |
|---|---|---|---|
| *.blogspot.com | 38 | 95 | All 38 open. 35 nofollow, 3 unreadable. Zero followed. |
| *.phorum.pl | 76 | 61 | Forum threads. No usable comment form. |
| *.websrvcs.com | 48 | 68 | Church media players. No comment form of any kind. |
Those three platforms were 162 rows, 16.2% of the list, and every row on each platform carried an identical rating. A DR 95 blogspot subdomain is not a DR 95 site. It is a free blog on a domain that belongs to Google, and the rating is a fact about blogspot.com, not about the page you were going to comment on.
Any vendor assembling a list by scraping for open comment forms will collect these platforms in bulk, because that is where open comment forms still are. So the inversion is not a coincidence of one bad list. It is what you should expect from any list built the same way.
Which makes the grouping test worth running first on anything you were sold. Checking a purchased list starts with exactly this: count distinct hosts, then look for blocks of rows sharing one rating. Both are spreadsheet work, both take minutes, and either can settle the question before you spend anything opening pages.
The general problem: a domain number cannot answer a page question
DR describes a domain’s backlink profile. Almost every decision in link prospecting is about a page:
- Can a link be placed here at all
- A property of one page and one template. Comments close when a post is archived; a theme update removes the form; a plugin switches every comment link to nofollow without touching the copy. The domain score does not move a point when any of that happens.
- What would the link carry
- Read from the links already on that page. Two pages on the same domain regularly disagree, because one runs the old template and one does not.
- Is this page about my subject
- A general news site has a food section. A photography blog has one post about accounting software. Judging either by the domain gets it wrong in both directions.
- Would anyone read it
- DR is a link-graph measure. It is not traffic, not engagement, and it says nothing about whether the page has a human audience.
When DR still earns its place
Deleting the column is the wrong response. DR is good at what it is for:
- Comparing whole sites when you are choosing between two publishers rather than two pages.
- Spotting an obviously dead domain. A DR 0 with no referring domains is usually exactly what it looks like.
- A tie-breaker at the end, once page-level checks have already produced a shortlist. Ordering ten qualified prospects by DR is reasonable. Ordering a thousand unqualified ones by DR is how you end up with 38 blogspot subdomains at the top.
- Client reporting, where a familiar number is easier to explain than a bespoke one, as long as it is not the number that made the decision.
What to put beside it
Four page-level facts, in the order they save you the most time:
- Can the page be read. 213 of our 1,001 rows could not be. Everything downstream is moot for those.
- Is the placement possible on this exact page. 818 rows had no comment form at all, whatever their domain scored.
- What the page already gives its commenters. Read from existing links, not assumed from the platform. See the guide on link attributes.
- What the existing comment thread looks like. The cheapest quality signal on the page, and the one that tells you what your link would sit next to.
Those four are the middle of a longer sequence. The full qualification order puts cheap spreadsheet rejections ahead of them and the three judgements only a person can make after them, which is the arrangement that keeps the expensive work for the rows that survived the cheap work.
Then bring DR back in as the tie-breaker it is good at being.
The honest limits of this finding
This is one list, one vertical mix, one date. It was a bought list, and bought lists are assembled by scraping for exactly the platforms that cause the inversion. A list you built yourself from editorial research will behave differently, and DR on that list may well correlate the way you expect.
What generalises is not the direction of the correlation. It is the reason: a hosted subdomain inherits its platform’s rating, and any list built by scraping open comment forms will be full of hosted subdomains. Check your own list for that pattern before you trust its DR column. Group by registered domain, look for large blocks of rows sharing one score, and you will know in a minute.
Checking it on your own list
The grouping test above is a spreadsheet job and worth doing regardless. If you want the page-level facts underneath it, the Backlink Prospect Checker reports all four on every URL you give it, with the evidence behind each one and an explicit Unknown where there was nothing to read. There is a sample report if you want to see the columns before signing in.