Brand Mention Taxonomy v1 is the classification layer used to distinguish between three different structures that can look similar in raw data.
The Gold Dataset calibration contains:
416 Explicit Brand Mention — Linked
328 Unlinked Brand Mentions
378 Linked Entity References
Together, those classes create:
1,122 taxonomy records
The explicit Brand Mention layer is:
744 explicit Brand Mentions
Within those 744 explicit mentions:
328 are unlinked
which equals:
44.1%
That percentage is one of the central findings of the Gold Dataset calibration.
It shows why a hyperlink count cannot be used as a direct substitute for Brand Mention count.
Brand Mention Taxonomy v1
The complete dataset documentation continues below in its original wording.
The three classes
Open section →02Dataset sectionUnlinked Brand Mention
Open section →03Dataset sectionLinked Entity Reference
Open section →04Dataset sectionWhy the taxonomy is necessary
Open section →05Dataset sectionHow the totals fit together
Open section →06Dataset sectionWhy 44.1% matters
Open section →The three classes
Explicit Brand Mention — Linked
This class contains cases where:
the brand is visibly named in the text
and
a link is present.
Example structure:
Volvo linked to a Volvo-related destination.
The defining feature is that the brand name is visible to the reader.
The link is additional structure.
Current count:
416
Unlinked Brand Mention
This class contains cases where:
the brand is visibly named in the text
but
no hyperlink is attached to that mention.
Example structure:
"I would consider Volvo for this type of car."
The brand is explicit.
The mention is unlinked.
Current count:
328
These records are important because a link-only analysis would miss them.
Linked Entity Reference
This class contains cases where:
a link points to a brand, company, product or related entity
but
the visible text does not explicitly name the brand in the same way as an explicit Brand Mention.
The anchor text may be descriptive.
It may contain a product name.
It may use generic wording.
The defining feature is that the entity is linked without a direct visible brand-name mention.
Current count:
378
These records matter because a link-only analysis could incorrectly count them as explicit Brand Mentions.
Why the taxonomy is necessary
Without taxonomy, two opposite errors appear.
Undercounting
If the analysis counts only linked mentions, it misses the 328 unlinked Brand Mentions.
Overcounting
If the analysis treats every outbound link as an explicit Brand Mention, it can incorrectly include Linked Entity References.
The taxonomy solves both problems.
It separates:
visible naming
from
hyperlink presence.
How the totals fit together
Gold Dataset v1 contains 794 outbound links.
But the Brand Mention taxonomy is not a one-to-one copy of the outbound-link count.
The public calibration contains:
416 linked explicit Brand Mentions
328 unlinked Brand Mentions
378 Linked Entity References
This creates:
744 explicit Brand Mentions
and
1,122 taxonomy records
The units are different.
An explicit Brand Mention is a visible naming event.
A Linked Entity Reference is an entity-linking event without equivalent visible naming.
That is why the totals should not be collapsed into one metric.
Why 44.1% matters
Among the 744 explicit Brand Mentions, 328 are unlinked.
328 divided by 744 equals:
44.1%
This means that in the Gold Dataset calibration, almost half of the explicit Brand Mentions did not depend on a hyperlink.
The finding is important because it shows that visible brand presence can exist independently of linking.
But the scope must remain clear.
This is a Gold Dataset calibration result.
It is not presented as a full-corpus prevalence rate.
A simple example
Consider three sentences.
Case 1
"Volvo is worth considering."
No link.
Classification:
Unlinked Brand Mention
Case 2
"Volvo is worth considering."
The word Volvo is linked.
Classification:
Explicit Brand Mention — Linked
Case 3
"See this model."
The text links to a Volvo-related destination, but the brand name is not visible.
Classification:
Linked Entity Reference
All three can contain a connection to the same brand.
But they are not the same type of evidence.
Why visible naming is its own signal
A visible brand name changes what the reader encounters.
The reader sees the company name directly in the text.
That is different from encountering:
a generic anchor,
a product name,
or a descriptive link.
The taxonomy therefore treats explicit naming as a separate structural property.
This does not mean explicit naming is always more valuable.
It means it should be measured separately.
Why linking is still useful
A link adds another layer of evidence.
It can connect the discussion to:
a company page,
a product,
a service,
or another entity.
That can help identify what the visible text refers to.
But the link itself does not determine whether the brand was explicitly named.
The taxonomy keeps those two properties separate.
Why this matters for Brand Mention reporting
A report that says:
"the brand received 500 links"
does not answer:
"how many times was the brand explicitly named?"
A report that says:
"the brand was mentioned 500 times"
does not answer:
"how many of those mentions were linked?"
Those are different questions.
Brand Mention Taxonomy v1 exists so both can be answered without conflating them.
Why this matters for AI Search research
AI systems can encounter both visible text and linked entities.
The current taxonomy does not claim that linked mentions, unlinked mentions and Linked Entity References have different effects on AI systems.
It establishes the categories needed to test that question.
Future experiments can compare:
linked explicit mentions,
unlinked explicit mentions,
and linked entity references
against AI-generated descriptions or recommendations.
That remains a future research question.
What this taxonomy does not claim
Brand Mention Taxonomy v1 does not claim that linked mentions are stronger than unlinked mentions.
It does not claim that unlinked mentions are more organic.
It does not claim that Linked Entity References are weaker evidence.
It does not claim that hyperlink presence automatically improves AI visibility.
It does not claim that 44.1% applies to the entire Full Research Inventory.
And it does not convert every outbound link into a Brand Mention.
The taxonomy has a narrower purpose:
to separate visible brand naming from hyperlink structure.
The practical takeaway
There are three different questions:
Was the brand visibly named?
Was the mention linked?
Was an entity linked without equivalent visible naming?
Brand Mention Taxonomy v1 keeps those questions separate.
That makes Brand Mention reporting more precise.