Quick answer

Evaluate a backlink in its context, not with a DR threshold. The deciding questions are: Does the reader of the source page have a reason to follow this link? Did the publisher choose the placement independently? Is the page accessible, indexable, and linked from the site? Is the commercial relationship disclosed and qualified? Does the link survive, and does it contribute to a useful audience or outcome?

A third party metric helps sort thousands of domains; it doesn't replace this review. Use a documented matrix, add absolute refusals for obvious schemes, and re-evaluate the link at 30, 90, and 180 days. For a paid link, the value should stay defensible even with rel="sponsored" and no ranking improvement.

Key takeaways

  • Relevance is a page, audience, destination relationship, not a general domain category.
  • Editorial independence and transparency separate a reference from a manipulated placement.
  • A broken link, one redirected to an irrelevant page, or a removed one shouldn't stay in the "acquired" total.
  • Repeated exact anchors, inconsistent neighborhoods, and network footprints raise risk; no single indicator proves a penalty.
  • The right decision can be accept, correct, qualify, monitor, or refuse, not just "good or bad."

The eight dimension matrix

Dimension 0 Max score Evidence
Relevance No connection to the task or audience 3 Page topic, reader need, destination
Real audience No sign of readers or an artificial page 3 Authorized Analytics, engagement, newsletter, brand search, estimated traffic with caveat
Editorial independence Text/anchor imposed only against payment 3 Validation process, author, right to critique
Contextual placement Footer, profile, or block with no justification 2 Sentence, section, depth, and neighboring links
Accessibility Page not reachable, soft 404, or non canonical 1 HTTP status, canonical, rendering, navigation
Verifiable traffic No evidence or an unverifiable screenshot 2 Source, period, countries, distribution by page
Qualification Hidden relationship or misleading attribute guardrail sponsored, ugc, nofollow, disclosure
Risk footprint none penalty 0 to 3 Ownership, templates, themes, vendors, anchors, dates

The sorting score is positive (0-14) minus risk (0-3). It is never automatic. A hidden commercial relationship, a paid link explicitly sold to manipulate Google, or an illegal site leads to refusal regardless of the total.

Nine step review procedure

1. Identify the origin

Does the link come from coverage, a partner, a directory, a user, a syndication, or a purchase? The same URL can be fine or risky depending on the relationship.

2. Read the whole page

Don't judge the domain inside a tool. Read the title, the author, the date, the sources, the sentence containing the link, the other links, and the next action. A strong page on an average domain can be more useful than a profile on a major outlet.

3. Test accessibility

Check HTTP response, canonical, rendering with no interaction, internal links to the page, and control tags. A page that's indexable today can become orphaned tomorrow; record the date.

4. Verify the relationship and the attribute

Google asks that paid links be qualified with sponsored and community links with ugc where relevant (Google, Qualify outbound links). The attribute doesn't replace user disclosure or the contract.

5. Examine relevance

Ask a simple question: with no SEO involved, would a reasonable publisher have added this link to help their reader? Note the exact usefulness: the source of a statistic, a tool, a definition, an example, or an action.

6. Check the footprint

Look for the same authors, anchors, templates, owners, catalogs, and publishing rhythms. Shared infrastructure isn't proof of spam. The signal becomes concerning when the only business model is selling links with no audience and no editorial line.

7. Classify the anchor

Natural anchors describe the destination or cite the brand. A concentration of exact match commercial anchors, especially in purchased content, calls for a review. Google doesn't publish an ideal ratio; don't invent one.

8. Record the expected outcome

Choose audience, referral, partnership, brand search, coverage, or learning. "Raise authority by 7 points" is a vendor proxy, not a business outcome.

9. Plan for survival

Check at 30, 90, and 180 days, then annually. Archive disappearance, changed attribute, redirect, canonical change, or context modification.

Ten classified cases

Case Likely decision Why
A journalist cites a SEOryon study with method Accept and track Independent, useful reference
A vendor lists its real partners Accept if accurate Visible relationship, legitimate discovery
Disclosed paid article, sponsored, sector audience Evaluate as media Audience value, not a ranking shortcut
Automatic syndication of coverage Deduplicate Same story, limited editorial control
Author profile on a recognized outlet Keep, low priority Real context but limited referral value
Relevant forum comment, disclosed brand Keep if useful Transparent participation; avoid self promotion
One-off exchange between complementary resources Review May help readers; refuse systematic scale
Industry directory with admission criteria Accept if a real public exists Sector discovery
Generic article sold with an exact followed anchor Refuse or correct Primary goal of manipulation
Network of hundreds of cloned multi topic sites Refuse No audience and high risk footprint

Site A publishes a signed interview with a security expert citing a SEOryon guide on data isolation. The page is reachable from the section, the audience is made of agencies, and the link serves a definition. Site B publishes "10 essential tools," with no author, nine commercial links, and an exact anchor. Both show DR 52.

With the scorecard: A scores relevance 3, audience 2, independence 3, context 2, accessibility 1, traffic 1, risk 0, total 12. B scores 1 + 0 + 0 + 1 + 1 + 0 minus 3 = 0. The difference comes from the document and the relationship, invisible in the number 52.

The backlink scorecard.csv lets you keep the evidence and the review date.

What the studies prove and don't prove

Ahrefs studied links pointing at 2,062,173 sampled sites starting from its January 2013 index. Nine years later, at least 66.5% were classified as lost; temporary errors and counting issues brought the broader category to 74.5% (Ahrefs, Link Rot Study). This is Ahrefs's crawl and definitions, on a historical cohort. The figure doesn't predict your link's survival; it shows the log must track the current state.

Large correlation studies often find an association between links and rankings, but don't answer the direction: a page can gain links because it's already visible. BuzzStream synthesizes fifteen studies and highlights their incompatible units (Is Link Building Dead?). Use this synthesis as a map to the sources, not as a universal causal effect.

A domain with a low DR, a foreign one, or one with no estimated traffic isn't automatically toxic. Most sites receive odd links without having created them. Act when you hold evidence of a controlled scheme, a manual action, or a serious history of manipulative purchases. Start by removing or qualifying the links you control.

Ahrefs's experiment on three pages, disavowing 3,476 links for four weeks then removing the file, observed a drop and a recovery in metrics (Ahrefs, Do Links Still Matter?). The test is too small, not randomized, and based on estimates to become a procedure. Don't replicate it on a client site.

Common mistakes

  • Setting a universal minimum DR.
  • Ignoring the page's context and looking only at the domain.
  • Refusing every nofollow link with no audience measurement.
  • Keeping vanished links in cumulative reporting.
  • Confusing correlation with causation.
  • Disavowing legitimate links on the recommendation of a toxicity score.
  • Forgetting rights, disclosures, and personal data in outreach.
  • Scoring precisely but with no archived evidence.

How SEOryon fits in

SEOryon can help produce content and assets that deserve references, spot relevant opportunities, and track organic performance. It cannot alone decide a publisher's independence or guarantee a link's value. The scorecard should stay reviewable by someone responsible who can refuse a placement even a well scored one.

Measurable exercise

Select fifteen real backlinks: five strong according to your tool, five weak, five random. Hide the third party metrics, have two people score the pages, then reveal the scores. The deliverable succeeds if every decision has evidence and disagreements are resolved by a rule, not by the reviewer's authority.

FAQ

A useful, contextual, transparent, and durable reference from a page aimed at a relevant audience. Its value can be editorial, commercial, or about discovery.

Is DR or DA more reliable?

These are different proprietary metrics. Use them as comparable filters within their own tool, never as a Google score or a final verdict.

Is a link from a page with no traffic worth anything?

Maybe, if the page is new, highly specialized, or serves as a reference. But the lack of audience and discovery reduces its immediate value; document the assumption.

After 30, 90, and 180 days, then at least annually or after a migration. Check existence, context, attribute, and destination.

Is an exact match anchor forbidden?

Not as a natural occurrence. The risk comes from a controlled, repeated scheme, especially a paid one. Google gives no ideal ratio.

Main sources

Method note

Method: the SEOryon scorecard is designed as a sorting tool, not a Google ranking factor. Sources verified 16 July 2026, translated and edited 22 July 2026.