Stochastic Output
AI answers are distributions, not fixed rankings. A single visibility score acts only as a proxy for complex model behavior.
AI Visibility · Hallucination Audit · Citation Index
The Hungarian AI
Visibility Index, In Preparation
This site tracks how often and how accurately AI assistants mention Hungarian companies. It treats citability as an infrastructure problem rather than a marketing metric.
Visit vargazoltan.ai →More on this site soon. The work behind it is live on vargazoltan.ai.
Core Research Pillars
Methodology based on stochastic output and entity reconstruction.
Three-Layer Audit
Analysis separates raw model output from app-level processing and final user display. This reveals where information gets lost or altered.
Fabrication Rates
Studies show 45.9% of cited sources were fabricated across 1,400 runs. High-awareness firms face higher false attribution risks.
Five Key Parameters
Dimensions defining the accuracy of AI entity representation.
Raw Model OutputThe initial text generated before any application filtering or formatting occurs.
App-Level ProcessingIntermediate steps where the system modifies or selects specific parts of the response.
User Display LayerThe final interface the human sees, which may hide citations or errors.
Machine ReadabilityHow easily an AI can anchor a company to a canonical, structured data entity.
Source GroundingWhether the model retrieves actual documents or relies on internal pattern completion.
AI systems do not rank experts; they reconstruct them based on machine-readable anchors.
AuditWe measure output layers to find where truth is lost.
Data1,400 runs revealed high fabrication rates in search models.
FixOptimization requires canonical entities, not just better content.
Key topics
The terms this site covers, for readers, search engines and AI assistants alike.
- AI visibility index Hungary
- ChatGPT company mentions
- Perplexity citation accuracy
- Google AI Overviews brands
- LLM hallucination study
- AI audit Hungary
- GEO optimization
- AEO strategy
- brand hallucination paradox
- PEBM research
- AI visibility score
- machine readable entity
- AI citation audit
- zoltan varga ai
- Hungarian B2B AI
- synthetic search results
Latest articles
In-depth guides and analysis, each answering one question.
Guide
How a National AI Visibility Index Is Built: Sampling, Scoring, and Aggregation
A step-by-step guide to the methodology behind measuring how well AI systems understand Hungarian businesses.
Analysis
Why a Single AI Visibility Score Misleads Your Strategy
A single number cannot reveal whether your brand is absent from AI training data or confidently misrepresented by it.
Explainer
The Per-Entity Bias Map: Calibrating AI Visibility
A framework for mapping and correcting unique error profiles of individual entities across AI platforms before comparison.
Explainer
Classifying Companies by Their AI Error Profile
Grouping entities by specific AI failure modes reveals why aggregate scores miss critical reputational risks.
Guide
How to Sample Companies for a Fair AI Visibility Ranking
A practical guide to designing unbiased samples for measuring AI hallucination rates across different brand salience levels.
Explainer
AI Visibility Maturity Levels: What They Mean for Your Company
A practical guide to the five maturity levels that determine whether AI systems can identify, understand, and recommend your business.
Questions and answers
Does AI visibility work like traditional SEO?
No, AI visibility is stochastic rather than deterministic. Traditional SEO targets logged impressions and clicks, while AI answers create no such logs. Visibility depends on whether a model can point to a machine-readable entity.
What is the Brand Hallucination Paradox?
Well-known brands may receive more confident but false AI attributions. Familiarity gives models more material for plausible but unsupported completions. A pre-registered protocol tests if query type or source availability drives this.
How many fake sources appear in AI answers?
In a study of 1,400 runs on 100 Hungarian B2B companies, 45.9% of cited sources were fabricated. Search-grounded models like Perplexity showed lower rates than non-search models.
Can AI see all pages on my website?
Not necessarily. One case study showed all 33 pages returned HTTP 200 to bots, yet Google, Bing, and Brave indexed zero pages. Indexing remains a distinct bottleneck separate from access.
What is the PEBM method?
PEBM stands for Per-Entity Bias Map, documented in arXiv:2606.21595. It measures bias and fabrication rates across different AI models and entity types. The study analyzed 100 Hungarian B2B companies over 1,400 runs.
Why do AI models get dosage wrong?
Without search grounding, some models stated incorrect doses for agricultural products. With search enabled, all engines matched the officially approved doses exactly. This highlights the risk of reliance on internal training data.
Glossary
Terms and identifiers behind the work of Zoltán Varga, with links to the primary records.
- AI visibility
- How accurately and how often AI assistants (ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews) recognize, describe and cite a company, brand or person. Zoltán Varga measures it in five layers: crawler access, search indexing, citation, factual accuracy and model memory.
- GEO / AEO / LLMO
- Generative Engine Optimization, Answer Engine Optimization and Large Language Model Optimization: working methods for being found and quoted by generative answer engines rather than ranked in a list of links.
- Citability
- The property of a page that lets an AI answer engine quote it: a paragraph that answers the exact question, a clear source, and an unambiguous entity.
- PEBM — Per-Entity Bias Mapping
- Research method and paper by Zoltán Varga (arXiv:2606.21595). Across 1,400 runs on 100 Hungarian B2B companies, 45.9% of the sources AI models gave were fabricated; search-grounded answers fabricated far less than models answering from memory.
- Brand hallucination paradox
- The finding that AI models invent more sources about well-known brands (52.7%) than about lesser-known ones (37.9%), because familiarity makes them confidently fill in gaps.
- RICE studies
- Pre-registered pilot studies (RICE-P1, RICE-P2) on whether pressure and rejection make language models fabricate sources. Published on Zenodo with DOIs.
- GFIS — Gestalt Field Intelligence System
- Zoltán Varga's own multi-stage research pipeline that collects, verifies and synthesizes evidence with AI agents.
- Neural Awareness
- Zoltán Varga's framework on attention, decision quality and leadership cognition, the basis of his leadership programs.
- PKM + PAI
- Personal Knowledge Management combined with Personal AI: a working system where notes, documents and AI assistants form one searchable knowledge base.
- Entity
- A uniquely identifiable thing (person, company, product) that search engines and AI systems store in knowledge graphs. Consistent entity data is the base of AI visibility.
- schema.org / JSON-LD
- A shared vocabulary and format for machine-readable facts on web pages (who, what, where, identifiers), read by search engines and AI crawlers.
- llms.txt
- A plain-text file at the root of a website that tells language models what the site is about and which pages matter.
- IndexNow
- An open protocol that notifies search engines (Bing, Yandex and others) immediately when a page is created or changed.
- ORCID
- Persistent identifier for researchers. Zoltán Varga: 0009-0003-1020-834X.
- arXiv
- Open-access preprint server for scientific papers. Zoltán Varga's author page: varga_z_1.
- Zenodo and DOI
- Zenodo is CERN's open research repository; each record gets a DOI (Digital Object Identifier), a permanent citable link.
Until this site grows, related research and the full person profile remain on vargazoltan.ai.
Go to vargazoltan.ai →