The Relevancy Rating™ score measures how strongly a backlink or citation reinforces the relationship between a linking page, a target page and the topic or category being assessed.
Relevance is part of how search engines and AI systems understand what a brand should be known for, and when it should be recommended. Relevancy Rating™ can support GEO and AiPR® activity by showing whether a placement is strengthening those associations.
Instead of relying on search engine data to infer relevance, the tool reads the linking and target pages directly, profiles the entire linking site, and scores the relationship across several independent signals.
The methodology combines natural language processing, entity analysis, vector embeddings, deterministic AI classification and mathematical verification, so every result is deterministic, consistent and explainable.
What the score helps you understand
Relevancy Rating™ is not designed to decide whether a placement is “good” or “bad” in isolation. It shows how strongly a backlink, citation or brand mention reinforces a specific topic, category or area of expertise.
A lower score does not mean a placement has no value. It may still support brand visibility, authority, referral traffic or wider awareness. It simply means the placement may be doing less to strengthen the specific topical relationship being assessed.
This is useful because different types of activity do different jobs. A Digital PR campaign may generate strong visibility and authority, while Hyper Relevancy Projects, targeted Digital PR or AiPR® activity may help strengthen more specific brand-topic associations over time.
Why traditional relevance tools get it wrong
Most relevance tools rely on Google’s index, third-party categories or surface-level page signals to infer what a page is about.
That can work for well-established URLs, but it becomes less reliable for:
- New pages
- Niche publishers
- JavaScript-rendered websites
- Content with limited index coverage
- Broad-interest domains that cover many unrelated topics
This creates a gap between what a tool thinks a page is about and what the content is actually saying.
Relevancy Rating™ takes a different approach.
Instead of relying on inferred signals, it reads the linking and target pages directly, profiles the wider linking domain, classifies relevance using a controlled AI framework, and verifies the assessment mathematically before producing a final score.
The result is a relevance score built from extracted content, semantic relationships and mathematical verification, rather than assumptions about how the page is categorised elsewhere.
How the tool assesses backlink relevance
Relevancy Rating™ uses a multi-stage pipeline to assess the relationship between the linking page, the target page and the topic or category being measured.
It combines direct content extraction, domain profiling, structured AI classification and vector-based verification to produce a score that is deterministic, consistent and explainable. We outline this in six steps below.
1. Direct content analysis
The tool fetches both the linking page and the destination page directly and in parallel, then processes both URLs through a multi-stage extraction pipeline.
The pipeline separates the main editorial content from navigation, adverts, boilerplate, templated modules and user-generated content, such as comment threads.
Where a page is JavaScript-rendered, the tool moves to a headless rendering stage. This allows rendered client-side pages and single-page apps to be analysed rather than being missed or misread from the initial source.
This means relevance is assessed against the article body and destination page content themselves, using the text a user, search engine or AI system is most likely to interpret.
2. Domain profiling
A single article doesn't always represent what a website is about. To measure site-level relevance, the tool profiles the wider linking domain rather than judging the backlink from one URL alone.
It crawls a representative sample of pages across the linking domain, prioritising core content and honouring robots.txt. Each eligible page is converted into a vector embedding.
These embeddings are combined into a length-weighted semantic centroid. In simple terms, rather than judging a backlink from a single article, Relevancy Rating™ builds a semantic fingerprint of the entire website at a broader semantic level.
The tool also measures the domain's topical coherence. This helps distinguish a specialist site, where the surrounding domain reinforces the topic, from a broad-interest publisher that may only cover the subject occasionally.
For example, a specialist gaming publisher linking to a games studio sends a different relevance signal from a general news site doing the same thing.
Because the tool assesses the site as a whole, it can also identify audience relevance. This can happen when the linking site shares an audience or locality with the target, even without a direct topical match.
3. Multi-layer relevance scoring
Each backlink is assessed across four independent relevance dimensions. These are scored separately before being combined into the final weighted result.
Domain Relevance
How closely the linking website's identity aligns with the target site's topic
Page Relevance
How closely the linking article matches the target page
Context Relevance
The topical relationship of the content immediately surrounding the link
Anchor Relevance
How descriptive and contextually relevant the anchor text is
4. Structured, deterministic AI classification
The language model doesn’t generate open-ended scores. Instead, it selects from a predefined set of relevance classifications in a single inference pass. The model runs at zero temperature, with a fixed seed and structured JSON output, so the classification process is controlled and repeatable.
The prompt is calibrated with worked reference examples, which anchor the model to the framework's rubric.
Each classification maps to a fixed score. The same inputs always produce the same labels, and the same labels always produce the same number. This keeps every evaluation consistent, repeatable and auditable, with no model "drift" between runs.
5. Mathematical verification
Each AI classification is independently checked against cosine similarity scores calculated from the vector embeddings.
Cosine similarity measures how close two semantic vectors are in the embedding space. In this methodology, it acts as a quantitative check on the language model’s interpretation.
Every relevance assessment is supported by two complementary forms of evidence:
- What the content means (linguistic classification)
- How it measures against the target (mathematical similarity)
Where the classification and similarity signals do not align, a set of guardrails are applied before the final score is produced to make the final assessment more comprehensive and robust.
6. Authority as a secondary signal
Topical relevance remains the primary ranking factor in the Relevancy Rating™ framework.
Domain authority is applied as a secondary signal through a stepped band lookup within the weighted formula. This allows highly authoritative sites to strengthen an already relevant backlink, without allowing authority to override poor topical fit.
In practical terms, a strong publication can add weight to a relevant placement, but it cannot make an irrelevant placement score highly on its own.
Can the tool assess mentions as well as links?
Although Relevancy Rating™ is primarily used to assess backlink relevance, the same page-to-page comparison can also be applied where no link is present.
This means the tool can be used to assess brand mentions, media coverage, citation opportunities and other page-level relationships.
For GEO and AiPR® activity, this is important because AI systems do not only rely on links. They also interpret the surrounding context of brand mentions, including the topics, categories, entities and sources connected to that brand.
Relevancy Rating™ helps show whether that context is reinforcing what a brand should be known for, even when the placement does not include a traditional backlink.
Transparent by design
The output includes individual component scores and an explanation of the signals that contributed to the final Relevancy Rating™ score.
This means the result can be understood and trusted. Users can see whether the score was driven by domain alignment, page similarity, local context, anchor text, authority, or a combination of those factors.
The core principle
Backlink relevance is not a single AI decision. It is evaluated as a staged technical process:
Combining these stages produces a far more reliable assessment than any single signal could on its own.
FAQs
Does a low Relevancy Rating™ mean a link is bad?
Does a low Relevancy Rating™ mean a link is bad?
No. A low score means the placement may not strongly reinforce the specific topic, category or page relationship being assessed.
But that does not mean the link has no value. Digital PR can still support brand visibility, authority, referral traffic and wider awareness.
Relevancy Rating™ helps show what a placement is contributing, so teams can understand where broader visibility is being built and where more targeted activity may be needed.
Should every Digital PR link be hyper-relevant?
Should every Digital PR link be hyper-relevant?
No. Digital PR links can serve different purposes, including authority, reach, awareness and reputation.
However, if the goal is to strengthen the relationship between a brand and a specific topic or category, relevance becomes more important.
Relevancy Rating™ helps separate broad visibility from topic reinforcement, so teams can see the role each placement is playing.
Is Relevancy Rating™ only for backlinks?
Is Relevancy Rating™ only for backlinks?
No. The Relevancy Rating™ tool can compare any two pages, whether there is a backlink, a brand mention or a wider citation opportunity.
This makes it useful for assessing off-site context, especially for GEO and AiPR® activity.
Can Relevancy Rating™ be used before outreach?
Can Relevancy Rating™ be used before outreach?
Yes. Relevancy Rating™ can be used to assess potential publications, pages or placement opportunities before outreach begins.
This can help teams prioritise opportunities that are more likely to strengthen the right topic associations, rather than only assessing placements after they have gone live.
How does Relevancy Rating™ support GEO?
How does Relevancy Rating™ support GEO?
GEO is partly about helping AI systems understand, cite and recommend a brand in the right contexts.
Relevancy Rating™ supports this by measuring whether a placement or mention strengthens the relationship between a brand, a topic and a trusted source.
How does Relevancy Rating™ support AiPR®?
How does Relevancy Rating™ support AiPR®?
AiPR® focuses on improving how a brand is understood across trusted sources, particularly around the topics, services and categories it wants to be known for.
Relevancy Rating™ helps measure whether placements and mentions are supporting those associations.
Does a low Relevancy Rating™ mean a link is bad?
No. A low score means the placement may not strongly reinforce the specific topic, category or page relationship being assessed.
But that does not mean the link has no value. Digital PR can still support brand visibility, authority, referral traffic and wider awareness.
Relevancy Rating™ helps show what a placement is contributing, so teams can understand where broader visibility is being built and where more targeted activity may be needed.
