FunkyMEDIA Research Lab was created because the agency had something most Brand Mention studies do not have:
a long-running publishing environment with its own history.
For approximately ten years, FunkyMEDIA has operated publishing infrastructure in which commercial discussions, recommendations, questions, product references and Brand Mentions accumulated across publisher-owned websites.
That material became the basis for the research now published on funkymedia.space.
The project did not begin with a random crawl of the internet.
It began with FunkyMEDIA's own publishing infrastructure.
The current research inventory contains 1,779,805 unique discussion URLs.
That makes the dataset valuable for a simple reason:
we know where the material came from, we know the infrastructure has existed over a long period, and we can separate the act of publication from the much longer period in which a Brand Mention can remain available on the public web.
From publishing infrastructure to research program
The first step was not to assume that 1.77 million URLs meant 1.77 million Brand Mentions.
A discussion URL is only the source unit.
The more difficult question is what actually appears inside the discussion.
Does the text explicitly name a brand?
Is the brand linked?
Is it named without a link?
Is the user asking about the brand or recommending it?
Does the wording repeat?
Does one mention contain only a company name, or does it connect that company with a product, problem, feature, experience or result?
Those questions turned the publishing inventory into a research program.
The current structure uses several layers.
The Full Research Inventory provides scale.
A 5,000-thread representative validation sample containing 8,708 posts was used to inspect the structure and chronology of the larger inventory.
Gold Dataset v1, containing 1,000 threads and 4,447 posts, provides the deep-analysis layer used for classification, repetition analysis and semantic research.
This separation is deliberate.
The project is designed so that a large inventory count does not silently become a claim about the meaning of every record inside it.
Why FunkyMEDIA publishes the findings
FunkyMEDIA is an AI Search and Brand Mentions agency.
That means the company works in an area where new terminology and strong commercial claims appear quickly.
AI visibility, brand recommendations, citations, mentions and entity context are increasingly discussed as if their meaning were already settled.
They are not.
FunkyMEDIA Research Lab exists to put a public evidence layer underneath that work.
Instead of publishing only conclusions, funkymedia.space publishes the research structure around them:
the main findings, the dataset layers, the graphs, the limitations and enough methodology to understand what was actually measured.
The goal is not to make every reader become a research analyst.
The goal is to make it possible to ask:
Where did this claim come from?
and receive a clear answer.
External brand context can persist for years.
A Brand Mention on a publisher-owned page behaves differently from content on the brand's own website.
The company may benefit from the visibility, but it does not own the page, domain, URL or surrounding editorial environment.
The publisher controls the infrastructure.
The brand benefits from the presence.
That makes Brand Mentions interesting not only as publication events, but also as persistent external brand assets.
A mention can remain available years after the original publication.
Its context can remain visible to search engines, researchers and AI systems.
The current research does not claim that this persistence causes AI recommendations.
It does show that external brand context can continue to exist long after publication and therefore deserves to be measured separately from publication volume.
The research is intentionally conservative.
The project does not treat every observed relationship as a verified fact about a company.
A Result relationship does not automatically prove causality.
A Co-Mention does not automatically prove competition.
Unique-like language does not prove organic authorship.
A Brand Mention does not automatically cause an AI recommendation.
And percentages from Gold Dataset v1 are not presented as if they described all 1,779,805 discussion URLs.
These boundaries are part of the research design.
They make the published conclusions narrower, but also easier to defend.
The research is published as separate, inspectable layers.
The outputs are separated so a reader can move from headline finding to underlying evidence without reading the project as one monolithic document.
unique discussion URLs
validation threads / posts
Gold Dataset threads / posts
Brand Mention taxonomy
analysis across explicit Brand Mentions
consideration, recommendation, warning, experience and comparison signals
semantic relationship records
unique Brand → Relation → Entity pairs
Brand Context Graph and related graph views
public case studies and research limitations
Rafał Cyrański
The methodology and research interpretation are authored by Rafał Cyrański.
FunkyMEDIA / FunkyMEDIA Research Lab
The research is published by FunkyMEDIA / FunkyMEDIA Research Lab.
BIGTECH IT Sp. z o.o.
The funkymedia.space service is operated by BIGTECH IT Sp. z o.o.
Those roles are intentionally separated. The research site is the publication layer. The agency develops and publishes the research. The operator provides the service infrastructure.