Large Language Models have changed the way brands are discovered, evaluated and talked about. BIKMA is the AI Visibility platform, built on a scientific method developed with the University of Pisa, in collaboration with the University of Insubria, and cited at The Web Conference, that measures how your brand appears to AI answer engines. It is the foundation of every project Fortop manages, and it is available self-serve at bikma.ai.
Until three years ago, the brand discovery funnel ran through Google and landed on your website. Today a growing share of that funnel never sees your website: it sees an answer generated by ChatGPT, Perplexity, Gemini or Claude, an answer that talks about your brand, describes its products and values, and compares it with competitors.
You didn’t write that answer. The models put it together from dozens of sources: your official content, third parties, news, reviews, social media, structured data. The model selects, summarises and synthesises, and in that synthesis it can say things that don’t match your truth.
For most of the brands we meet, this is a complete blind spot: nobody measures what machines say about them, nobody knows their generative Share of Voice, nobody checks whether the models cite the brand accurately. It is a gap that can be closed, and now is the right time, before it becomes a structural cost.
Every tool measures model answers. What sets BIKMA apart is how it builds the questions they are measured on. If the question set is arbitrary, every number derived from it is arbitrary. That is why BIKMA does not rewrite keywords as questions: it reconstructs the market’s real questions from scientific literature and public conversation, samples each intent with multiple phrasings, and weights each question with the Prompt Probability Index, a declared probability rather than an observed volume.
The scope, the competitive set and the question clusters are defined with you at project kick-off. From there, the measurement becomes repeatable over time and comparable across markets and models.
The market’s real demand, explicit and latent, segmented into branded and non-branded clusters and weighted with the Prompt Probability Index. It is the starting point for understanding where your brand should be visible and isn’t yet, and for whom: the audience clusters behind the demand.
The brand’s presence in model answers across a set of key questions, against a defined competitive set: how often you are cited, in what narrative position, and with what weight compared with competitors.
The sources that feed the models about your brand, classified by type and domain authority. Are they your official content, or third-party sources, news, social media, reviews, retailers, trade publications? The generative knowledge graph becomes explicit and mappable. Models systematically favour third-party sources over the brand’s own website.
The tone of the answers: positive, neutral or critical, and which attributes the model’s judgement focuses on (price, quality, ethics, innovation, sustainability). It is the brand’s public face, filtered through the machines.
How closely what the models say about your brand matches what the brand officially states on its own website. It measures the risk of generative misinformation, increasingly critical for pharma, regulated food and finance.
BIKMA (Brand Insight Knowledge Mapping Analytics) grew out of a research collaboration with the University of Pisa, with contributions from the University of Insubria. The model was cited at The Web Conference in San Francisco, the leading international scientific conference on the Web, and is anchored in the latest scientific literature (2024-2026) on information behaviour and generative models. We keep updating it alongside university research: the extension that measures visibility in AI answer engines is our teams’ most recent work.
We have applied it to more than 200 enterprise projects for brands in pharma, food, luxury, manufacturing and banking. It is the same method we bring to large companies as a project managed from start to finish.
with the University of Pisa, with contributions from the University of Insubria.
San Francisco. Paper cited at the leading international scientific conference on the Web.
enterprise. Bayer, Lactalis, Chiesi, Barilla.
and introduction of the Prompt Probability Index.
A European pharmaceutical multinational discovered that models were citing its active ingredient alongside contraindications superseded by two regulatory updates: the generative knowledge graph was stuck in 2022. Six months of targeted work on sources and fact checking brought fact alignment from 67% to 94%.
An Italian FMCG food brand had a generative Share of Voice of 12% against a direct competitor at 31%, despite a larger market share. The BIKMA analysis revealed that the competitor held nine key sources, including trade publications, nutrition blogs and retailer product pages, that the brand did not consider central to its editorial plan.
An Italian luxury brand was cited with 78% positive sentiment, but with significant narrative drift: the brand’s heritage was systematically associated with a city where the company had never operated. It was a chain of third-party citations built on a Wikipedia entry that had never been corrected. A single correction realigned the knowledge graph in eight weeks.
If you need a structured action plan (content, SEO/GEO/AEO, advertising, knowledge graph management, third-party source relations, ongoing fact checking), we build a tailored project and run it as a managed service, from start to finish. It is the path chosen by most enterprise brands.
If you prefer to run continuous monitoring in-house, or you are an agency looking after your own clients, BIKMA is available self-serve at bikma.ai: a free trial to get started right away, then a subscription with real-time data. It is also the path for large companies that choose the platform but have their agency of record run it.
Tell us your scope: markets, brands, category, goals. We’ll start from there to build the managed project that fits best. If you’d rather see the platform first, the bikma.ai free trial is just a click away.