Introduction
A vendor sheet lands in your inbox, and before you read a single URL, your eyes go straight to one column. DR. Or DA. You sort the sheet by that number, high to low, and you start picking from the top, working your way down until the budget runs out. It feels efficient. It feels like due diligence. You just paid for the domains with the biggest number sitting next to them, so how could this go wrong?
Everybody does this. That is exactly the problem.
Here is the uncomfortable truth: the number you trust most when you evaluate backlink quality is not a Google signal. It never was. It is a third-party estimate built by a tools company, sold to you as a proxy for authority, and it can be manufactured by the very vendors who are selling you links off it. The metric that was supposed to protect you as a buyer has become the exact metric being used against you. You are not the only one who has learned to sort by that column. So has every vendor trying to get a placement past you.
Link quality was never a number on a spreadsheet. It is decided by a handful of signals that DA, DR, and Authority Score cannot see at all: relevance, traffic direction, placement context, anchor history, and the neighborhood a domain keeps. None of these live in a proprietary formula. All of them are checkable, if you know to look. The buyer who learns to read those five signals stops overpaying for authority that does not exist, and starts buying links that actually move rankings and keep moving them past the next core update.
By the time you finish this article, you will have the complete 5-signal evaluation framework we run on every domain before a client link goes live. You will know exactly what each signal catches that the DR column hides from you, why vendors are counting on you never checking it, and how to run this evaluation at scale without turning link buying into a full-time forensic exercise. Let’s start with the number itself, and where it actually comes from, because most buyers have never asked.
The Number You Trust Was Never a Google Signal
Three Companies, Three Formulas, One Score You Treat as Gospel
Domain Authority, Domain Rating, and Authority Score sound like they should mean roughly the same thing, and the industry treats them as if they do, using them interchangeably in Slack threads, client reports, and vendor sheets. They don’t come from the same place, and none of them come from a search engine. DA is a Moz metric. DR belongs to Ahrefs. Authority Score is SEMrush’s version of the same underlying idea. Three separate companies, three proprietary formulas, and three different scores for the exact same website, because each company built its own model of what “authority” should look like based on its own link index, its own crawl frequency, and its own weighting logic.
Not one of these numbers lives inside Google’s algorithm. Google has never confirmed, endorsed, or even acknowledged DA, DR, or Authority Score as ranking factors, because Google simply does not use them. They are external estimates, reverse-engineered after the fact, from data each company crawled on its own schedule with its own index. Google’s own John Mueller has said as much publicly, more than once, and the industry keeps buying links off the number anyway.
So why do agencies, buyers, and vendors treat these scores like gospel if Google never touches them? Because at scale, someone has to filter a firehose of options down to something usable. When a link vendor is sitting on a database of hundreds of thousands of domains, they cannot manually vet every single one before it goes on a sheet. They need a fast, repeatable way to cut that pile down to a shortlist that clears some minimum threshold a client will accept. A DR filter does that job well. It is quick, it is quantifiable, and it lets a vendor say “everything on this list is DR 40 plus” without reading a single page of content on a single domain.
That is the real function of the score. It is a sorting mechanism built for people working at volume, not a certification of quality, and definitely not a guarantee that a link will do anything for a client’s rankings.
Here is the line that reframes everything that follows: a filter is not a verdict. A high DR can tell a vendor which domains are worth a second look before they build a sheet. It cannot tell you, the buyer, whether a link on that domain will actually help a client rank, whether it carries any real editorial weight, or whether it is safe from the next update. Treating a filtering tool as a final judgment is exactly how experienced, careful operators end up buying bad links with total confidence, because the number gave them permission to stop looking any further.
What a High DR Actually Measures (And How It Gets Faked)
The Number Is Real. The Authority Is Not.
Let’s be precise about what DR is actually measuring, because the confusion starts right here. DR measures the size and diversity of the backlink profile pointing at a domain. That’s the entire calculation. It counts links, weighs them by the authority of the linking domains inside Ahrefs’ own index, runs that through a proprietary formula, and outputs a single number between zero and one hundred.
Now look at everything that calculation leaves out, deliberately, one item at a time. Relevance is missing, because DR has no concept of whether the linking site has anything to do with your client’s industry, market, or audience. Traffic is missing, because DR does not check whether a single real visitor lands on that domain in a given month, or a given year. Placement context is missing, because DR cannot tell a contextual link inside a well-read article apart from a link buried in a footer next to fifty unrelated outbound links. Anchor history is missing, because DR does not flag a domain that has already been hammered with exact-match commercial anchor text by every other buyer who came before you. Neighborhood is missing, because DR does not check who else that domain links out to, or what kind of company it keeps across its own content.
Put plainly, DR does not know whether the site is relevant to your client, whether a real human ever visits it, or whether your link is sitting three inches away from a casino affiliate site and a payday loan blog on the same page. It counts links, runs them through a formula, and hands you a number that looks precise and feels earned. That is the entire job it does, and it is the only job it was ever built to do.
Here is the second half of the problem, and it is the part vendors do not put on the sheet next to the score. If a score is built purely by counting links pointing at a domain, then anyone who is capable of building links at a domain can inflate that score, deliberately and predictably. This is not a theoretical risk that only affects careless buyers. It happens constantly, at scale, as a business model. A vendor buys an expired domain that once carried real, earned authority in a legitimate niche. They rebuild a thin site on top of the old structure, point a wall of links at it from their own private network to pump the DR back into an impressive range, and list the domain for sale the moment the number clears whatever threshold moves inventory fastest. You paid for a DR 60. You received a rebuilt shell with a manufactured score, a manufactured backstory, and no real audience behind any of it.
“The metric that was supposed to protect you is the exact metric being used against you.”
This is why sorting a vendor sheet by DR alone is not a shortcut, it is a trap that rewards whoever is best at gaming the metric, not whoever actually has the best link to offer a client. The vendors who understand this best are the ones building their entire inventory around exploiting it.
The 5-Signal Link Evaluation Framework
What We Check Before a Single Link Goes Near a Client
Stop opening the DR column first. If the score cannot see relevance, traffic direction, placement, anchor history, or neighborhood, then those are exactly the things you need to check yourself before a domain earns a place in front of a client. Here are the five signals we run on every prospective placement, in order, and the specific failure each one is built to catch that a big number on a sheet will never reveal.
Signal 1: Traffic Trend
What it checks is not a domain’s current traffic number sitting in isolation, it is the direction that traffic has been moving over a meaningful window of time, typically six to twelve months of history. Why it matters is that a growing traffic line is strong evidence of a living site that is actually earning attention from real readers and, by extension, from Google’s own ranking systems, because organic traffic and organic rankings tend to move together. A flat line sitting near zero, no matter what the DR says directly above it, is the signature of a domain being propped up on paper alone, with a backlink profile built to impress a spreadsheet rather than an audience. The failure this signal catches is the rebuilt expired domain: high DR, impressive-looking metrics at a glance, and an audience of essentially nobody once you actually pull the traffic history. The pass condition is organic traffic that trends upward, or at minimum holds steady, over that window, not a single spike from one viral post three years ago and not a flatline quietly propped up underneath a big authority score.
Signal 2: Topical Relevance
What it checks is whether the domain is genuinely, consistently about your client’s world, not adjacent to it, not vaguely related to it in a stretch, actually built around it over a real publishing history. Why it matters is that relevance is context Google reads directly from the content and the surrounding link graph, and it is context the DR column ignores completely, treating a link from a finance site and a link from a pet food blog as functionally identical as long as the authority number matches. The failure this signal catches is the content mill: a site that publishes about crypto on Monday, weight loss on Wednesday, and dog food on Friday, because it exists purely to sell placements to whoever pays that week, not to serve any coherent topic or any real reader. The pass condition is a consistent topical identity that genuinely overlaps with your client’s niche, built over time through actual editorial decisions, not assembled hastily right before the domain went up for sale.
Signal 3: Placement Context
What it checks is where the link physically sits on the page and what surrounds it on that same page. Why it matters is that an editorial mention woven naturally into real, substantive content carries a completely different weight than a link dropped into a resource dump at the bottom of a page, because one is something a reader would actually encounter mid-article and the other is something they would scroll straight past without a second glance. The failure this signal catches is the footer link or the sidebar “recommended resources” list, buried among a dozen other unrelated outbound links that exist purely to host paid placements for whoever is currently buying. The pass condition is a link that lives inside genuine editorial content, in context, where a real reader following the article’s argument would naturally land on it as part of the flow, not stumble onto it in a list.
Signal 4: Anchor Profile
What it checks is the anchor text history already built at the domain, meaning what the existing inbound links to that site currently say, not just what your own anchor text will say once you add to it. Why it matters is that a profile already stuffed with aggressive, exact-match commercial anchors is a manipulation signal in its own right, one that Google’s own systems are built to detect, and it is a signal that transfers directly onto your client’s site the moment your link joins that same profile. The failure this signal catches is a domain that already looks manipulated to anyone who runs a basic backlink report, meaning it is one algorithmic update away from real scrutiny, and your client’s link is sitting inside that exposure whether you knew it or not. The pass condition is a natural, varied anchor profile, branded terms, naked URLs, generic phrases like “click here” or “learn more,” with no aggressive stuffing of commercial keywords across the majority of inbound links.
Signal 5: The Neighborhood
What it checks is who else the domain links out to across its own published content, not just who links in. Why it matters is that a link is judged, in real and measurable part, by the company it keeps, and a domain that will link out to anyone who pays is a domain sitting one click away from spam, low-trust affiliate sites, and content Google’s spam systems already treat with active suspicion. The failure this signal catches is the domain that has quietly become a link farm hub, technically legitimate-looking on the surface with a clean design and real-seeming articles, but functionally a switchboard for paid placements pointing outward in every direction to whoever is currently buying. The pass condition is outbound links that point to real, credible sites operating in legitimate, relevant spaces, evidence that the domain still has editorial standards governing what it links to rather than an open checkout page disguised as a blog.
None of these five signals show up anywhere in the DR column, and every single one of them decides whether the link you are about to buy actually works for a client or quietly does nothing while the invoice still clears.
How to Run This at Scale Without Lowering the Bar
Where AI Actually Belongs in Link Evaluation
Here is the objection you are already forming as you read this: checking all five signals, by hand, on every domain across every vendor sheet you receive in a given month, is not realistic at any real scale. You are right, and this is exactly where most of the market gets AI backwards.
The prevailing approach treats AI as a replacement for the fundamentals rather than an accelerant for them. Feed a tool a list of domains, let it spit out a composite score, buy whatever clears the threshold, done in an afternoon. That is not scale, that is automating bad decisions faster than a human team ever could make them manually, and it produces the exact same manufactured-authority problem the industry already has, just with an AI-shaped rubber stamp sitting on top of it now instead of a human one.
Our position is different, and it is the one we build into every agency partner account we manage. We use AI to do the heavy lifting on the fundamentals, at a scale no human team could realistically match working link by link. It scans traffic patterns across thousands of domains at once, flags relevance mismatches between a domain’s actual content history and a specific client’s niche, surfaces anchor profiles that already carry manipulation risk before a single dollar changes hands, and maps the outbound neighborhood of every domain under consideration, all in minutes instead of the weeks it would take a team working manually through the same volume.
Then, and this is the part that actually matters, a human vets it. AI flags the pattern across the full volume of candidates. A person with real link-building judgment, someone who has watched these patterns play out across hundreds of live campaigns, decides whether that pattern actually fits this specific client, this specific niche, and this specific risk tolerance the client is willing to accept. Neither step replaces the other, and removing either one breaks the process.
AI speed. Human intelligence.
The practical takeaway is simple, and you can apply it starting with the next vendor sheet that lands in your inbox. Stop asking what the DR is. Start observing the niche, the traffic direction, the relevance, and the placement context instead, in that order. The next time a vendor points at a big number on a sheet as the reason to buy, ask them the questions that number was never built to answer: what does the traffic trend look like over the past year, what does this domain typically publish about, where exactly does the link sit on the page, what does the existing anchor profile already carry, and who does this domain link out to elsewhere on the site. If they cannot answer those questions clearly and specifically, you already have your answer about the link, regardless of what the score says.
Conclusion
The number was never the quality. It was just easy. Sorting a vendor sheet by DR takes ten seconds and feels like due diligence, which is exactly why it became the default across the industry, and exactly why it keeps producing the same outcome for buyers who trust it: a shelf full of expensive links that do nothing measurable for a client’s rankings once the invoice clears.
Operators who keep buying on the score keep overpaying for authority that does not actually exist, and they keep inheriting risk they cannot see until an update exposes it publicly in a client’s traffic graph. Operators who evaluate on the five signals instead, traffic trend, topical relevance, placement context, anchor profile, and neighborhood, end up buying fewer links overall, spending less per placement that actually counts toward rankings, and building link profiles for clients that hold up over time instead of collapsing at the next core update.
If you want to see how this evaluation actually runs before a single link goes live for a client, and how we do it at scale without lowering the bar anywhere in the process, the walkthrough shows the exact process we run across every agency partner account we manage today. It is not a model we are asking you to outsource to us blindly and hope for the best. It is a model built to learn from and replicate directly inside your own operation, on your own accounts, with your own team.
Book a free Agency SEO Process Walkthrough. See the link evaluation behind our agency partner accounts. No pitch. Just the process.
[Book Your Free Agency SEO Process Walkthrough]
