# aiprofil.hu > AI Profile makes company facts consistent and machine-readable. Optimizes entity data, schema.org, and llms.txt for AI visibility. This site defines an entry-level package to make company facts consistent and machine-readable for AI systems. It solves the infrastructure problem where models reconstruct entities rather than rank them. ## 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 - machine readable facts - entity optimization - schema.org for AI - llms.txt guide - AI citation - brand hallucination - GEO strategy - AEO optimization - LLMO tactics - AI profile package - AI visibility audit - AI profil - gépi olvasható adatok - AI láthatóság - szintézis optimalizálás ## Questions and answers ### How does AI visibility differ from traditional SEO? AI visibility is stochastic, meaning results vary with each query, unlike deterministic SEO rankings. Models reconstruct entities based on available data rather than following a fixed algorithmic ladder. ### What is the PEBM study finding regarding citations? The PEBM study found that 45.9% of cited sources in AI answers were fabricated overall. High-awareness firms saw a 52.7% fabrication rate compared to 37.9% for low-awareness firms. ### Why might well-known brands face more hallucinations? Well-known brands may receive more confident but false attributions due to the brand hallucination paradox. Familiarity gives models more material for plausible but unsupported completions. ### What does an AI Profile package include? An AI Profile package ensures company facts are consistent and machine-readable for AI systems. It covers entity data standardization, schema.org implementation, and llms.txt creation. ### Can AI systems track clicks on their answers? No, an AI answer creates no logged impression or click in the traditional sense. Measurement requires analyzing the three output layers from raw model text to the user interface. ### What is the cost of measuring AI visibility? Cost depends on scope: how many AI engines are tested, how many questions are asked and how often each question is repeated. A focused measurement can be run for a small budget; a full audit with repair work is scoped individually. ## 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 Goes Into a Machine-Readable Company Profile](https://aiprofil.hu/articles/machine-readable-company-profile/): A machine-readable company profile is a structured set of data that explicitly defines a business entity, its people, its offerings, and its relationships using standards like schema.org and JSON-LD. It serves as the foundational identity layer that allows AI systems to anchor a brand without guessing, ensuring the business appears in automated decisions and agent-driven transactions. - [Schema.org for Companies and Experts: What to Use](https://aiprofil.hu/articles/schema-org-for-companies-and-experts/): Companies must implement Organization schema with sameAs links to authoritative profiles, while experts require Person schema linked to their employer. These core types, combined with Service or Product markup and FAQPage for specific queries, create the machine-readable identity layer that AI systems need to anchor content. - [sameAs Links: Connecting a Brand's Scattered Identities](https://aiprofil.hu/articles/sameas-and-identity-links/): sameAs links declare that distinct web addresses represent the exact same real-world entity, allowing AI systems to merge scattered profiles into one coherent knowledge graph. Without these explicit connections, AI treats a website, a LinkedIn page, and a registry entry as separate, unverified fragments rather than a single trusted source. Implementing these links ensures that citations, reputation signals, and factual data flow seamlessly across the entire digital footprint. - [llms.txt Explained: Purpose, Format and Real Weight for AI Visibility](https://aiprofil.hu/articles/llms-txt-explained/): llms.txt is a lightweight text file designed to guide AI crawlers toward your most relevant content, functioning similarly to robots.txt but for large language models. While early adopters implement it to signal priority pages, current data shows major AI bots rarely fetch it, making it a supportive infrastructure element rather than a primary driver of visibility. - [How to Stop AI from Confusing Your Brand with Similar Names](https://aiprofil.hu/articles/entity-disambiguation-guide/): To stop AI from confusing your company with similarly named entities, you must explicitly map name collisions, assign unique distinguishing attributes to your specific organization, and repeat those same identifiers consistently across every digital platform. This creates a machine-readable signal that allows retrieval systems to separate your brand from others during the generation process. - [Should You Create Your Own Wikidata Entry?](https://aiprofil.hu/articles/self-created-wikidata-entries/): No, companies and experts should not create their own Wikidata or Wikipedia entries because these platforms strictly enforce notability rules that require independent, third-party sources. Self-created entries are frequently deleted, whereas durable structured authority comes from earning independent coverage and maintaining a consistent, machine-readable identity layer on your own domain. - [Checklist: Keeping Company Facts Consistent Across the Web](https://aiprofil.hu/articles/entity-consistency-checklist/): You keep company facts consistent by establishing one canonical fact sheet that defines your legal name, leadership, and core offerings, then enforcing this data across your website schema, public registries, and third-party profiles. Inconsistencies in names, dates, or roles create ambiguity that causes AI systems to generate errors or hallucinate incorrect information about your organization. - [Does Structured Data Shape AI Answers Without Being Cited?](https://aiprofil.hu/articles/structured-data-indirect-effect/): Yes, structured entity data from sources like Wikidata and schema.org influences AI answers by entering the model's training corpus, where it becomes part of the parametric memory without requiring explicit citation. This indirect integration shapes the model's understanding of relationships and facts, allowing AI systems to reference entities accurately even when the original structured source is not linked in the final output. - [Why Wikipedia Weighs Heavily in What AI Models Know](https://aiprofil.hu/articles/wikipedia-weight-in-model-knowledge/): Wikipedia matters because its structured, canonical entity definitions are heavily weighted during large language model training, embedding specific facts directly into the model's parameters. This creates a parametric advantage where recognized entities are recalled from memory rather than retrieved from external sources. Entities lacking this encyclopedic structure depend entirely on real-time retrieval systems to be known by the AI. - [Making an Individual Expert's Profile Machine-Readable](https://aiprofil.hu/articles/machine-readable-expert-profile/): An individual expert must replace generic biographical text with structured data, including schema.org markup, a dedicated llms.txt file, and unique research identifiers like DOI or Wikidata Q-numbers. These elements allow AI systems to distinguish the expert from namesakes and ground their responses in verified facts rather than training-data hallucinations. Without these anchors, the expert remains a 'ghost entity' that models either ignore or misrepresent with fabricated credentials. - [Canonical Presence: The Anchor AI Needs for an Entity](https://aiprofil.hu/articles/canonical-presence/): Canonical presence is the condition where an entity possesses a single, authoritative, machine-readable anchor point in the digital field, such as a stable knowledge graph identifier and consistent structured data. An entity needs this anchor because without it, AI systems cannot distinguish between verified facts and fabricated details, leading to inconsistent or hallucinated representations across platforms. - [Machine-Readable Profile FAQ: Schema, llms.txt, Wikidata](https://aiprofil.hu/articles/machine-readable-profile-faq/): Start by creating a Wikidata entry to establish a unique identifier, then implement schema.org markup on your official site to link that identifier. A single knowledge base entry is insufficient because AI systems require cross-layer verification across knowledge graphs, entity recognition, and linguistic structures. The process typically requires three to six months of sustained effort to overcome regional infrastructure gaps.