# ailathatosag.hu > Measuring and repairing how AI assistants describe companies. A quarterly Hungarian AI Visibility Index and workshops for marketers are in preparation. This site is about measuring, auditing and repairing how AI assistants describe companies. A quarterly Hungarian AI Visibility Index and workshops for marketers are in preparation. ## Author - Zoltán Varga — AI visibility researcher and AI ecosystem architect, Budapest: https://vargazoltan.ai/ - English profile: https://zoltanvarga.ai/en/about/ - ORCID: https://orcid.org/0009-0003-1020-834X - arXiv author: https://arxiv.org/a/varga_z_1.html - Zenodo: https://zenodo.org/search?q=owners%3A1653886 ## Publications - Per-Entity Bias Mapping for AI Visibility: Why Brand Mentions Require Entity-Specific Calibration (arXiv:2606.21595): https://arxiv.org/abs/2606.21595 - AI Visibility in the Generative Era: A Manifesto for Entity-Level Measurement (DOI 10.5281/zenodo.20285424): https://doi.org/10.5281/zenodo.20285424 - Quality Challenges Do Not Escalate Source Fabrication in Large Language Models (RICE-P1) (DOI 10.5281/zenodo.20474050): https://doi.org/10.5281/zenodo.20474050 - Rejection-Induced Confabulation Escalation — Pilot Study 2 (RICE-P2) (DOI 10.5281/zenodo.20509755): https://doi.org/10.5281/zenodo.20509755 ## Key topics - AI visibility - AI audit Hungary - brand hallucination - AI citation repair - GEO optimization - AEO strategy - LLMO services - Hungarian AI Index - AI visibility workshop - AI search audit - AI visibility mérése - AI láthatóság - márka hallucináció - AI audit Budapest - AI kereső optimalizálás - AI láthatóság javítása ## Questions and answers ### How does AI visibility differ from traditional SEO? AI visibility is stochastic, meaning answers change based on context and lack logged clicks. Unlike SEO, which targets deterministic rankings, this field focuses on how models reconstruct entities from available data. ### What is the Hungarian AI Visibility Index? It is a quarterly report measuring how AI assistants describe Hungarian companies. The index tracks citability and fabrication rates across models like ChatGPT and Perplexity. ### Why do famous brands get more fake AI citations? This is known as the brand hallucination paradox. Familiarity gives models more material for plausible but unsupported completions, leading to confident errors. ### Can you fix AI hallucinations about my company? Yes, through audits that identify missing machine-readable anchors. Fixes involve restructuring content so models can point to canonical entities rather than guessing. ### What data supports the PEBM methodology? The Per-Entity Bias Map analyzed 1,400 runs across 100 Hungarian B2B companies. It found that 45.9% of cited sources were fabricated overall. ### Do AI crawlers always see your website correctly? Not always. Technical access may return HTTP 200, but indexing bottlenecks can prevent search-based models from finding specific pages. ## Related - AI visibility expert page: https://vargazoltan.ai/ai-lathatosagi-szakerto/ - AI visibility audit: https://vargazoltan.ai/audit/ - Research: https://vargazoltan.ai/kutatas/ ## Articles - [What an AI Visibility Audit Actually Measures Beyond SEO](https://ailathatosag.hu/articles/what-an-ai-visibility-audit-measures/): An AI visibility audit measures per-user context variability, layered output integrity, and entity accuracy rather than search rankings and click-through rates. Unlike SEO audits which track deterministic impressions, this audit quantifies the statistical probability of a brand appearing across diverse user prompts and platforms. It focuses on whether the AI correctly identifies and cites the entity, not whether a user clicked a link. - [The Three-Layer Output Problem in AI Answers](https://ailathatosag.hu/articles/three-layer-output-problem/): The same AI model produces different brand descriptions because the output passes through three distinct layers: raw model generation, application-level processing, and final user rendering. Brand mentions frequently vanish during safety filtering or text truncation in the middle layer, creating a gap between what the model knows and what the user sees. This structural variance explains why identical queries yield inconsistent results across different platforms. - [How to Design the Query Set for an AI Visibility Audit](https://ailathatosag.hu/articles/designing-the-audit-query-set/): An effective AI visibility audit requires a stratified query set covering direct brand, category, comparison, and problem-based questions. This specific mix determines the audit result because AI responses are stochastic and vary significantly by context, unlike deterministic search engine rankings. - [The Golden Dataset: A Stable Reference for AI Visibility Audits](https://ailathatosag.hu/articles/golden-dataset-ai-visibility/): A golden dataset in AI visibility measurement is a curated collection of query-answer pairs where human experts have pre-defined the correct brand references. It acts as a stable ground truth against which stochastic model outputs are scored, replacing unreliable single-number scores with consistent, drift-detectable metrics. - [Checklist: Facts to Prepare Before an AI Visibility Audit](https://ailathatosag.hu/articles/pre-audit-fact-inventory-checklist/): To prepare for an AI visibility audit, a company must compile precise legal identity data, full profiles of key personnel, detailed service descriptions, and a map of media mentions and name collisions. This ground truth allows the auditor to distinguish between actual brand authority and noise in stochastic AI models. Without these specific facts, the audit cannot measure the distribution of visibility across different query types and platforms. - [Which AI Visibility Claims Survive Rival Hypotheses?](https://ailathatosag.hu/articles/rival-hypotheses-ai-visibility-claims/): Most common AI visibility claims fail under rival hypothesis testing because they confuse correlation with causation or mistake platform artifacts for strategic drivers. Only claims supported by controlled experiments distinguishing between sampling noise, structural pipeline loss, and genuine causal mechanisms remain valid. You must treat every metric as a probability distribution rather than a fixed point to survive these tests. - [Does Ad Spend Make a Brand Appear in AI Answers?](https://ailathatosag.hu/articles/pay-to-appear-hypothesis/): Companies with higher ad budgets often appear more frequently in AI answers, but this correlation is driven by increased content production and domain authority rather than direct payment mechanisms. Only specific platforms like Perplexity offer explicit sponsored answers, while most AI models rely on training data frequency and relevance. - [Why Clicks Fail as a Metric for AI Visibility](https://ailathatosag.hu/articles/why-clicks-fail-as-ai-metric/): Click-through rate is a poor measure of AI visibility because modern AI answers often resolve user queries without requiring a website visit, creating a zero-click environment where success equals no click. Traditional search console data suffers from survivorship bias, capturing only the minority of interactions that result in further navigation while missing the majority of brand mentions that satisfy users instantly. - [What a 12-Week AI Visibility Repair Plan Looks Like](https://ailathatosag.hu/articles/twelve-week-ai-visibility-repair-plan/): A 12-week AI visibility repair plan begins with a baseline audit to identify specific failure modes, followed by three phases: fixing technical infrastructure, generating citable content, and establishing cross-platform authority links. Success depends on the client actively updating personal profiles and securing external mentions, as AI systems require multiple independent sources to recognize an entity. - [Which AI Visibility Interventions Have Evidence?](https://ailathatosag.hu/articles/evidence-for-ai-visibility-interventions/): There is currently no published causal evidence proving that specific GEO, AEO, or LLMO interventions directly improve entity citation in generative AI. Existing research relies on theoretical frameworks like Retrieval-Augmented Generation rather than controlled before-and-after trials. Brands should focus on building verifiable third-party knowledge infrastructure while awaiting controlled attribution studies. - [AI Visibility Audit FAQ: What Can and Cannot Be Promised](https://ailathatosag.hu/articles/ai-visibility-audit-faq/): An AI visibility audit cannot guarantee specific ranking positions or eliminate hallucinations because large language models operate on independent parametric memory and retrieval layers. The audit measures current entity infrastructure gaps and provides a repair plan, but final results depend on the client executing cross-platform content and link-building tasks. Success requires a 12-week collaborative sprint, not a one-time technical fix. - [How to Read AI Visibility Audit Results: A Practical Guide](https://ailathatosag.hu/articles/reading-ai-visibility-audit-results/): To interpret AI visibility audit results, you must analyze the 25 individual dimension scores rather than relying on a single aggregate number. High scores in identity and personal signal dimensions often outweigh technical perfection, as AI systems prioritize verifiable human entities over anonymous content. A low total score usually indicates missing structural evidence of who stands behind the organization, not a failure of the business itself.