A Link Is Not Always a Brand Mention

The starting point: 794 outbound links

Gold Dataset v1 contains:

• 1,000 threads
• 4,447 posts
• 794 outbound links

A conventional link audit could stop there.

It could count destinations, domains and anchors and then describe which entities received links.

But that would answer only one question:

Where did the discussion point?

Brand Mention research asks another:

Was the brand actually visible in the language?

That difference is small at first glance and important in practice.

What happens when the links are classified

The Gold Dataset taxonomy separates three structures:

• 416 Explicit Brand Mentions — Linked
• 328 Unlinked Brand Mentions
• 378 Linked Entity References

The first class contains visible brand naming and a matching link.

The second contains visible naming without a matching link.

The third contains a link to an entity-owned destination without necessarily showing the brand name in the visible text.

That means a link audit and a Brand Mention audit overlap, but they do not produce the same result.

Linked Entity References are the key difference

Consider a sentence that says:

"you can see the full offer here"

with "here" linked to a company website.

The destination identifies the company.

The visible text does not.

That is a useful entity reference, but it is not the same thing as a sentence that visibly says the brand name.

This distinction matters because users, search systems and language models can process visible wording differently from destination URLs.

A Brand Mention is evidence that the brand itself appears in the text.

A Linked Entity Reference is evidence that the text points toward the entity.

Those are related signals, not identical ones.

The opposite case also matters: Brand Mentions without links

The Gold Dataset contains 328 unlinked Brand Mentions.

These are explicit brand references with no matching outbound link.

Within the 744 explicit Brand Mentions in Gold Dataset v1, unlinked mentions represent 44.1%.

That figure belongs to the calibrated dataset and should not be generalized to the full research inventory.

Its significance is straightforward:

if a monitoring system looks only at backlinks, it will miss explicit brand visibility that exists in text without a link.

A user can recommend a company without linking to it.

A user can ask whether a brand is trustworthy without linking to it.

A user can describe an experience with a brand without linking to it.

The absence of a hyperlink does not make the Brand Mention disappear.

Why this changes the way Brand Mentions should be measured

There are two common measurement errors.

The first is undercounting:

counting only linked references and missing unlinked Brand Mentions.

The second is overcounting:

treating every link to a company-owned destination as if the brand had been explicitly named.

The taxonomy solves both problems by separating the units.

That makes link data more useful, not less useful.

It becomes possible to ask two clean questions instead of forcing one metric to answer both:

How often was the brand explicitly named?

and

How often did the discussion point to the brand or entity?

What 794 outbound links can and cannot tell us

The 794-link figure is still valuable.

It tells us that the Gold Dataset contains a substantial link layer.

It creates a bridge between Brand Mention research and traditional link analysis.

But the link count does not, by itself, tell us how many explicit Brand Mentions occurred.

For that, the visible text has to be classified.

This is why the research does not convert outbound links directly into Brand Mention volume.

The two signals are kept side by side.

Why this matters for AI Search research

AI systems can encounter several forms of brand evidence:

the visible name,

the linked destination,

the surrounding descriptive language,

the entities and products mentioned nearby,

and the wider context of the page.

The current study does not claim that one of those signals is always more important for AI systems.

It does show that those signals can exist separately.

That is a necessary distinction before anyone tries to study how AI systems use brand context.

If naming and linking are mixed together at the measurement stage, later conclusions about AI visibility become harder to interpret.

The practical takeaway

A backlink is not proof that a brand was explicitly mentioned.

An explicit Brand Mention does not require a backlink.

If a report uses only one of those signals, it is measuring only part of the external brand context.

The better approach is to keep them separate:

explicit naming

and

entity linking

Once they are separated, it becomes possible to study how each behaves over time, across different contexts and eventually in relation to AI-generated answers.

What this study does not claim

The study does not claim that links are unimportant.

It does not claim that unlinked mentions are automatically more valuable.

It does not claim that Linked Entity References have no value.

It does not claim that explicit naming causes AI recommendations.

The finding is narrower:

a link and a Brand Mention are different analytical units.

That distinction is the result.

Related research

Not All Brand Mentions Are Equal
The broader taxonomy behind linked, unlinked and entity-reference classes.

The Repetition Problem
Why volume also needs to be separated from repeated wording.

Brand Mentions Before the Decision
How explicit brand visibility appears before recommendation.

Gold Dataset v1
The deep-analysis dataset containing the 794 outbound links.

Research Limitations
The boundaries that apply to the findings.