airanking.hu / Articles / 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.

By Zoltán Varga2026-11-056 min read

Defining the Population and Sampling Scope

The construction of an AI visibility index begins with defining the universe of potential subjects. In the Hungarian context, this involves identifying a large set of public domains representing the active business sector. A recent measurement covered 20,957 distinct domains to establish a representative baseline for the national economy. This scope ensures the data reflects the actual digital footprint of small and medium-sized enterprises rather than a curated selection of high-performing websites.

Not every domain in the initial list yields usable data for the final index. Technical failures, slow load times, or blocked access prevent the audit system from retrieving content. From the total pool, 16,899 domains successfully passed the initial technical checks to receive a full audit. Of these, 16,614 underwent the complete 25-dimension evaluation required for accurate scoring. The remaining 285 were assessed using a reduced seven-dimension framework due to data limitations.

The distinction between the total universe and the audited sample is critical for interpreting the results. The index does not claim to measure every single website in the country but rather provides a statistically significant snapshot of the market. With a coverage rate of 79.28% of the total domain universe receiving the full 25-dimension audit, the data offers a robust view of the current state of AI readiness among Hungarian businesses without claiming absolute completeness.

This sampling strategy avoids the bias of only analyzing popular or well-known companies. By casting a wide net across the entire domain registry, the index captures the reality of the average business, which often lacks sophisticated digital infrastructure. The goal is to measure the structural capacity of the national economy to interact with AI agents, not to rank individual brand performance. This approach ensures the resulting index serves as a diagnostic tool for the sector rather than a competitive leaderboard.

Collecting and Auditing Web Presence

Once the sample is defined, the system performs a deep crawl of each website to gather raw data. This process differs fundamentally from traditional search engine optimization audits. Instead of checking for keyword density or backlink counts, the crawler analyzes how an AI agent perceives the page structure and content. It examines whether the site provides clear entity definitions, structured data schemas, and verifiable business information that machines can parse without human interpretation.

The audit evaluates the technical accessibility of the site as a prerequisite for any further analysis. Approximately 15% of the examined domains faced technical blocks that prevented the AI from accessing the content entirely. These issues range from server errors to restrictive robots.txt files. Without resolving these basic technical hurdles, the remaining 24 dimensions of the audit cannot be measured, rendering the site effectively invisible to AI-driven decision-making systems.

For the majority of sites that pass the technical check, the audit focuses on content quality and structure. The system looks for explicit statements about what the company does, who owns it, and where it operates. It checks for the presence of FAQs, service descriptions, and contact details that are formatted for machine reading. The analysis reveals that 85% of the problems found are not technical but relate to the lack of structured, citable information within the content itself.

This phase of the process identifies the gap between human readability and machine interpretability. A website might look professional to a human visitor but fail to provide the specific data points an AI agent needs to make a recommendation. The audit flags missing structured data, ambiguous service descriptions, and a lack of external verification signals. These findings form the basis for the subsequent scoring, highlighting where businesses must shift from visual design to data-centric communication.

Scoring Dimensions and Entity Calibration

The core of the index relies on a 25-dimension scoring model designed to mimic the logic of modern AI agents. These dimensions are grouped into five thematic areas: identifiability, offer clarity, credibility, technical foundation, and local context. Each dimension contributes to a final score out of 100, reflecting how well the AI can understand and trust the business entity. This granular approach allows for a nuanced diagnosis rather than a simple pass-or-fail result.

Identifiability is often the most deficient area in the current dataset. It measures whether the website contains machine-readable identifiers that link the digital presence to a real-world organization. Many sites lack structured data markup or clear connections to official registries. Without these signals, AI systems struggle to distinguish the business from others or to verify its existence, leading to a low score regardless of the content quality.

Credibility and offer clarity assess the depth and reliability of the information provided. The system checks if the content offers concrete answers to potential questions rather than vague marketing slogans. It also evaluates the presence of trust signals such as citations, case studies, or references to external authoritative sources. Research indicates that AI models exhibit a bias toward known entities, meaning smaller businesses must work harder to establish this digital trust through explicit evidence.

The scoring model accounts for the specific challenges of the Hungarian language and local context. Global AI models are trained primarily on English data, creating a structural disadvantage for non-English content. The audit dimensions specifically measure how well the site supports Hungarian language processing and local reference structures. This ensures the index reflects the reality of the domestic market, where businesses must overcome linguistic barriers to be visible to global AI agents.

Aggregation by Sector and Benchmarking

After scoring individual domains, the data is aggregated by industry sector to reveal broader trends. The classification uses automated signals to group businesses into categories such as technology, healthcare, finance, and general trade. This aggregation allows for the comparison of AI readiness across different parts of the economy. It highlights which sectors are ahead in adopting structured data practices and which are lagging behind in digital evolution.

The results show significant variation between sectors, though none have reached the highest maturity levels. Technology and education sectors tend to score higher, with averages around 49 and 48 respectively, due to their existing focus on digital content. However, even these leading groups fall into the 'AI-Lagging' category, scoring below 60. The general business category, representing the largest volume of sites, averages just 33.63, indicating a widespread lack of AI-optimized infrastructure.

Sector-specific benchmarks serve as a guide for targeted improvement rather than a judgment of business success. A lower score in the financial sector, for instance, reflects the strict compliance requirements and complex nature of the data rather than poor performance. The index helps stakeholders understand that a business in the 'AI-Invisible' range is not failing commercially but is simply not yet equipped for the new era of agent-based discovery and procurement.

These aggregated insights enable the creation of sector-specific development programs. Policymakers and industry associations can use the data to design training and support initiatives tailored to the specific gaps in each industry. For example, the tourism sector might need help with structured location data, while legal services require better citation practices. The index transforms raw data into actionable intelligence for economic development.

Update Cycle and Responsible Publication

An AI visibility index is not a static report but a dynamic tool that requires regular updates to remain relevant. The methodology includes a defined update cycle to track progress over time. As AI models evolve and new standards emerge, the scoring dimensions must be adjusted to reflect the changing capabilities of these systems. This ensures the index remains a reliable barometer of the national digital landscape.

Responsible publication of the index avoids naming individual companies as winners or losers. Publicly ranking specific businesses could create reputational risks and mislead stakeholders who might interpret a single audit as a definitive judgment. Instead, the focus remains on patterns, trends, and aggregate data. This approach protects privacy while still providing valuable insights into the overall health of the digital ecosystem.

The index is designed to support continuous improvement rather than one-time certification. Businesses are encouraged to view their score as a starting point for ongoing optimization. As they refine their content and structure, they can re-audit their presence to measure progress. This iterative process aligns with the rapid pace of AI development, where what is considered 'ready' today may change tomorrow.

Future iterations of the index will incorporate feedback from the business community and advancements in AI technology. The goal is to establish a standard for AI-readiness that becomes widely adopted across the economy. By maintaining a transparent and rigorous methodology, the index fosters trust and encourages businesses to invest in the necessary digital infrastructure. This long-term perspective ensures the index remains a valuable asset for years to come.

AI-assisted: drafted with a language model from the author's research notes, then edited and checked before publication. Author: Zoltán Varga, AI visibility researcher and AI ecosystem architect, Budapest.