FunkyMEDIA opens public research lab for AI search visibility

FunkyMEDIA has launched a public research program in Poland to measure how brands appear inside AI search and online conversations. The lab says its new dataset and calibration work separate linked mentions, unlinked mentions, repetition and semantic context to give a more precise view of brand visibility.

Why it matters: - AI-generated search is changing how users discover companies, which makes brand visibility harder to measure with a single count. - FunkyMEDIA Research Lab is trying to separate the different signals behind a brand appearance, including links, explicit mentions, repetition and semantic context. - The project aims to create a clearer measurement layer before anyone draws stronger conclusions about how mentions affect AI recommendations.

What happened: - FunkyMEDIA launched FunkyMEDIA Research Lab as a public research program focused on how brands appear, are discussed and gain context across online conversations. - The research is published at FunkyMEDIA.space and draws from a cleaned inventory of 1,779,805 unique discussion URLs. - The inventory comes from selected exports of FunkyMEDIA’s historical and operational publishing infrastructure. - Rafał Cyrański is the founder of FunkyMEDIA and the author of the research methodology.

The details: - Gold Dataset v1 includes 1,000 threads and 4,447 posts as a manual calibration layer. - The calibration layer identified 744 explicit Brand Mentions. - Of those mentions, 416 were linked and 328 were unlinked. - 44.1% of explicit Brand Mentions in the Gold Dataset appeared without a matching link. - The same dataset found 63 exact formulaic mentions and 681 unique-like mentions under the current repetition method. - FunkyMEDIA says “unique-like” does not mean organic. - FunkyMEDIA says the calibration results should not be treated as prevalence estimates for the full 1.7 million URL research inventory. - The semantic calibration layer contains 536 relationship records and 276 unique Brand → Relation → Entity pairs. - Those relationships include products, services, problems, use cases, features, results, locations and co-mentions. - FunkyMEDIA Research Lab publishes datasets, methodology, case studies, graphs, limitations and research interpretations publicly. - The launch extends FunkyMEDIA’s work in AI Search and Brand Mentions, with FunkyMEDIA.space serving as the public evidence and research layer of the broader FunkyMEDIA ecosystem. - FunkyMEDIA is a Polish AI Search and Brand Mentions agency founded in 2010.

Between the lines: - The lab is making a methodological argument: a brand mention is not one uniform signal. - A brand can be named with a link, named without a link, or referenced through a destination URL without visible brand text. - Mention strength, evidence, repetition and surrounding language can all change what a mention means. - The public dataset approach also signals caution. FunkyMEDIA is separating measurement from causal claims about recommendation outcomes. - Rafał Cyrański said the distinction matters because brands are increasingly understood through relationships, context and evidence distributed across many sources.

What's next: - FunkyMEDIA will continue publishing datasets and research interpretations through FunkyMEDIA.space. - The lab is positioned to expand its evidence base as AI Search keeps reshaping discovery behavior. - Future work will likely focus on refining how mentions, relationships and context are measured before broader claims are made about brand visibility.

The bottom line: - FunkyMEDIA is betting that AI search visibility needs a finer-grained research model than simple mention counts. The company’s public lab is designed to show how brands are represented, not just whether they appear.