Building a Europe AI Can See: Hungary’s National AI Profile

This research is part of the Building A Europe AI Can See series, which also includes National AI profiles of Austria, Belgium, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary.
Read also about Hungary Flagship companies.

Why Hungary, and Why Now

The five audited pages are the ones that speak for the Hungarian state: the government overview page, and the ministries of science and technology, economy and energy, justice, and education and children’s affairs. They sit on one platform, kormany.hu. These are the pages an AI system is most likely to be asked about when someone wants to know who is in the government, what a ministry has announced, or which deadline applies. The audits measure whether those pages are readable, verifiable and current to a machine.

The pages are live, time-sensitive public information. The audited queries carry a ministry meeting held on 1 September 2026, a nursery needs survey with an October 22 deadline, a licence plate replacement deadline of November 30, 2026, and the 2027 round of the Bursa Hungarica scholarship. The government overview page lists the prime minister, deputy prime ministers and ministers by name.

The result splits in two. On the dimension where Greece was weakest, Hungary is strong: every one of the five pages has a single H1, a logical heading hierarchy and H2 sub-sections. The same five pages declare almost nothing. Each carries one JSON-LD block, a bare WebPage. Four of the five carry no date signal of any kind.

The Profile in One View

The five Hungarian pages average 55.8 on the AI Retrieval Index. No institution reaches a B. Two score Grade C (Fair) and three score Grade D (Poor). The spread between best and worst is 6 points.

Dimension averages across the five, with the Greek averages from the Greece and German National AI Profile averages it reported for reference:

The main finding: Hungarian government pages are structured and extractable, but not declared or dated. Structure averages 98.0, 40 points above Greece and 24 above Germany. Schema averages 23.0 and freshness 9.0, both below Greece and Germany. Entity Clarity is 55 at all five institutions. Hungary finishes 6.4 points ahead of Greece on the overall index because structure outweighs the weaker schema and freshness scores.

Other measured indicators:

• Every page has exactly one H1. The single-H1 check, the H2 sub-section check and the logical nesting check (H1→H2→H3) pass at all five. Balanced heading density fails at two: the Government page and the Ministry of Education and Children’s Affairs.
• One JSON-LD type across the set. Each page carries one valid block, a WebPage. Organization, Person and Article schema were not detected at any of the five.
• Knowledge-graph anchoring (sameAs links) scored 0/100 at all five. Entity Connectivity and E-E-A-T Density scored 0 at all five, and no expertise, experience, authoritativeness or trustworthiness signals were detected on any page.
• Freshness scored 0 at four institutions. Only the Ministry of Education and Children’s Affairs scores above zero, at 45. A structural decay alert was raised at the other four.
• Meta description fails at four of five. Only the Government page passes it.

How the Five Engines Read the Hungarian Pages

The research scores each page against five engines, each with a defined role: Perplexity (The Researcher), OpenAI / ChatGPT (The Generalist), Claude (The Semanticist), Google Gemini (The Integrator) and Microsoft Copilot (The Knowledge Connector). The audit also gives each engine an estimated citation percentage.

Gemini is the weakest engine for Hungary, as it was for Greece. Its average score is 39.6, identical to the Greek average, and its average citation estimate is ~20.2%. The audit flags the same two gaps at all five institutions: no Article or Person JSON-LD, which leaves E-E-A-T unverifiable, and a missing Person and Organization pairing. A missing canonical or meta description is flagged at four, and the Government page is flagged for image alt-text coverage of 54%. The two highest Gemini scores belong to the two pages that carry something the other three lack. The Government page (42) passes the meta description check. The Education page (45) is the only one with date signals. The audit does not attribute Gemini’s scores to those checks.

Claude scores highest and is cited at ~46% at all five. Claude averages 79.4, 18.2 points above its Greek average of 61.2, and its citation estimate is ~46% at every institution against ~24.0% in Greece. The audit raises the same two flags on every page: no author or Person schema, and low contextual disambiguation. It raises no heading-hierarchy flag, because the H1 and nesting checks pass everywhere.

ChatGPT scores 75–77 with a ~50% citation estimate at all five. The audit flags weak entity signals at every institution and, at the same time, finds content depth and structure optimised for semantic retrieval. ChatGPT is 12.8 points above its Greek average of 63.8.

Copilot returns the same result for every page. It scores 66 with a ~46% citation estimate at all five institutions, 10.0 points above its Greek average of 56.0. The audit flags the missing Article and Person schema at every page and recommends adding Person and Organization schema together.

Perplexity is the most varied engine and the least moved by structure. Scores run from 45 to 53 and citation estimates from ~46% to ~59%. The Government page is the only page where the audit raises no Perplexity issue (“core citation signals detected”), and it carries the highest Perplexity citation estimate (~59%) while scoring 51. The Education page scores highest (53, ~48%). The other four pages are flagged for a missing meta description. Perplexity averages 47.8, 1.4 points above its Greek average of 46.4.

Structure lifted three engines and left two unchanged. Against Greece, Claude gained 18.2 points, ChatGPT 12.8 and Copilot 10.0. Perplexity gained 1.4 and Gemini 0.0. The structural fix moved the engines that read hierarchy. It did not move the engines the audit describes as dependent on structured data, authorship signals and freshness (Perplexity, Gemini).

What the Five Audits Reveal

Structure is the strongest dimension in the set. Structural Integrity averages 98.0 (range 95–100). Three pages score 100 on all five checks. The Government page and the Education page score 95 and fail only the balanced-heading-density check. The fragmented-intent check passes at all five.

Extractability is solid, and breaks on tables. Scores run from 75 to 80 (average 76.0). Word density (300+ words), list structures and internal link density pass at all five. Table / structured data fails at all five. Image alt-text coverage fails at one page, the Government page, which has 19 images missing alt text (54% coverage). It passes at the other four.

Entity clarity is identical across the set. Every page scores 55 (Grade C). Primary entity tokenization and Open Graph Title pass at all five. Article / content schema, Author / Person schema and Organization schema fail at all five.

Schema and metadata is the second-weakest dimension. Scores run from 20 to 35 (average 23.0). The robots, canonical and Open Graph checks pass at all five, and one valid JSON-LD block is present on each page: five blocks in total, all WebPage. The meta description check fails at four pages. The Government page passes all five metadata checks and still scores only 35 (Grade D). The other four score 20 (Grade F).

Freshness splits the set into four and one. The Science, Economy, Justice and Government pages score 0 on every freshness check: no JSON-LD dateModified, no machine-readable date, no publish and modified pair, no recorded update and no content-age signal. A structural decay alert reads “No date signals found — content age is unverifiable.” The Education page scores 45 (Grade D). It carries a date detected through text or meta tags, passes the 180-day and 12-month content-age checks and shows update-frequency signals. It fails JSON-LD dateModified, the publish and modified pair and the content-updated check. The audit recommends moving its dates into JSON-LD.

Trust and authority signals are zero. Entity Connectivity is 0, KG Anchoring is 0 and E-E-A-T Density is 0 at all five institutions. The E-E-A-T Signal Breakdown shows 0 signals each for Expertise, Experience, Authoritativeness and Trustworthiness everywhere. The Entity Relationship Map shows no link between Person, Article and Organization on any page, and no knowledge-graph anchor.

Entity graph stability is moderate, and it is not built on markup. Entity Graph Stability averages 68.0 (range 63–70), rated Moderate at all five. The Greek average was 32.2. Primary entity scores 20/20 and Clarity 15/15 at all five. Secondary scores 18–20 out of 20 and Co-occurrence 10–15 out of 15. The Schema sub-score is 0/20 and Queries is 0/10 at every institution. Hungary reaches a moderate stability score on page content alone. Semantic Richness averages 50.0 (range 46–52).

Queries are plentiful, and interface labels are among them. The audits detect 127 queries across the five pages (25, 34, 15, 26 and 27). 101 are flagged eligible, and no estimate exceeds 55%. Navigation and utility text is read as topics. “Kiemelt tartalmak” (featured content) and “fő navigáció” (main navigation) appear at all five. “Elérhetőségek” (contacts), “Dokumentumtár” (document library), “adatkezelési tájékoztató” (privacy notice) and “közérdekű adatok” (public-interest data) appear at four. On the Government page, “minisztérium megtekintése” (view ministry) merges with minister names and titles across line breaks.

Entity detection finds the platform, not the ministries. The same entity, https://kormany.hu, is detected as an Organization brand entity at 68% confidence on all five pages, with no schema. No topic clusters were detected on any page. In the displayed entity labels, names are cut at the first accented character: “Tarr Zolt”, “Nemzeti Innov”, “Az Igazs”, “Bursa Hungarica Fels”, “Senyei Gy”. The audit displays them in this form.

Institution Profiles

Government of Hungary (Kormányzat)

Website audited: https://kormany.hu/en/government
Role: The government overview page. The Inspector reads the page as being about “Kormányzat” (the government). The detected content covers the prime minister, deputy prime ministers and ministers by name, the government spokespersons, the list of ministries and the document library.
Audit result: 57 – Grade C (Fair) · Structure A · Depth B · Schema D

Strongest findings:

• It is the only page in the set that passes all five metadata checks: JSON-LD present, meta description, canonical, Open Graph tags and robots. Schema & Metadata of 35 is the highest in the set.
• It is the only page where the audit raises no Perplexity issue. Perplexity scores 51 with a ~59% citation estimate, the highest Perplexity citation estimate in the set.
• Entity Graph Stability is 70 (Moderate), with Secondary 20/20, Co-occurrence 15/15 and Clarity 15/15.
• It yields the most queries of the five (34), and 29 are flagged eligible. The queries are the names and titles of the government’s members.
• ChatGPT scores 77 (~50%) and Claude 79 (~46%).

Weaknesses:

• Freshness is 0. There is no dateModified, no machine-readable date and no update signal. A structural decay alert is raised.
• 19 images lack alt text (54% coverage). It is the only page in the set where the alt-text check fails, and Gemini flags it.
• Balanced heading density fails.
• The only JSON-LD is a WebPage block. There is no Organization, Person or Article schema, and Entity Connectivity, KG Anchoring and E-E-A-T Density are 0.
• Gemini scores 42 with a ~16% citation estimate, the lowest Gemini citation estimate in the set.
• Several ministers appear as bare name queries (Kapitány István, Kátai-Németh Vilmos, Pósfai Gábor) with 45% eligibility and no role attached. Three queries are multi-line strings that merge a “view ministry” label, a name and a title.

What it means for AI systems: This is the page an AI is most likely to consult to answer “who is in the Hungarian government?” It names every member, but a machine cannot confirm any of those names as a Person with a role, cannot tell when the roster was last updated and has no declared publisher to check. A government roster is the kind of page where currency decides correctness, and this page gives no date signal.

Ministry of Science and Technology

Website audited: https://kormany.hu/en/government/tudomanyos-es-technologiai-miniszterium
Role: The Tudományos és Technológiai Minisztérium. The detected content covers the minister, Tanács Zoltán, and his biography; the Innovation Voucher Program; a new programme for doctoral students; the Hungarian Startup University; a HUF 100 million call under the MTMI Innovation Workshop; new functions of the public data portal; the launch of the mihivatal.gov.hu government site; Tanács Zoltán’s remarks on autonomous vehicles; technology and robotics talks in Hong Kong; criminal complaints in the Gondosóra case and over the land registry system; the suspension of the mandatory extension of Poszeidon EKEIDR; and the November 30, 2026 plug-in hybrid licence plate deadline.
Audit result: 54 – Grade D (Poor) · Structure A · Depth B · Schema F

Strongest findings:

• Structural Integrity is 100. Every structure check passes, and the alt-text check passes.
• Claude scores 80 (~46%) and ChatGPT 77 (~50%).
• The detected queries read as concrete news and programmes: the Innovation Voucher Program, the doctoral programme, the MTMI call, the government site launch. 19 of 25 are flagged eligible.
• Semantic Richness is 52.

Weaknesses:

• No meta description. Schema & Metadata is 20 (F), and Perplexity (45, ~46%) flags the missing summary anchor.
• Freshness is 0, although the page carries a dated deadline (November 30, 2026) and a “two months left” notice.
• Entity Graph Stability is 63, the lowest in the set, with Secondary 18/20 and Co-occurrence 10/15.
• The minister is named in several news headlines and has a biography query, yet no Person schema exists. The audit shows E-E-A-T Density at 0.
• Gemini scores 37 with a ~21% citation estimate.
• Detected entities are truncated fragments (“Nemzeti Innov”, “Impresszum Adatkezel”, “Nemzeti Jogszab”).

What it means for AI systems: The page covers fast-moving subjects: innovation funding, technology policy, public data. The content is readable and headline-rich, and nothing in the markup tells an engine who the minister is, who publishes the page or when it was written.

Ministry of Economy and Energy

Website audited: https://kormany.hu/en/government/gazdasagi-es-energetikai-miniszterium
Role: The Gazdasági és Energetikai Minisztérium. The detected content covers the minister, Kapitány István, and his biography; changes to the liquidity credit products of the Széchenyi Card Program; an announcement on securing Hungary’s energy supply; and exploration for hydrocarbons at three of seven sites announced in February.
Audit result: 54 – Grade D (Poor) · Structure A · Depth B · Schema F

Strongest findings:

• Structural Integrity is 100, and the alt-text check passes.
• The detected entity “Energetikai Miniszt” (the ministry name, truncated) carries 80% confidence, the highest entity confidence in the Hungarian set.
• Entity Graph Stability is 70 (Moderate), with Secondary 20/20 and Co-occurrence 15/15.
• Claude scores 80 (~46%) and ChatGPT 77 (~50%). 12 of 15 queries are flagged eligible.

Weaknesses:

• No meta description, and Schema & Metadata is 20 (F).
• Freshness is 0. The page announces energy policy and credit-product changes, which are time-sensitive, and gives no date signal.
• Semantic Richness is 46, the lowest in the set, and the page yields the fewest queries (15).
• The minister’s biography query is classed as a Tool-Query at 49%, and no Person schema exists.
• Programme names (Kisfaludy Program, Magyar Falu Program) are detected as co-occurrence signals at 48–56% with no schema.
• Gemini scores 37 (~21%) and Perplexity 45 (~46%).

What it means for AI systems: Energy and credit-policy changes are the pages where a stale answer causes harm. The page tells AI what changed and gives no machine-readable date for it.

Ministry of Justice

Website audited: https://kormany.hu/en/government/igazsagugyi-miniszterium
Role: The Igazságügyi Minisztérium. The detected content covers the minister, Görög Márta; consultations on Hungary’s accession to the European Public Prosecutor’s Office (EPPO) and a meeting with the European chief prosecutor; talks with the National Intellectual Property Office, the Hungarian Association for the Protection of Industrial Property and Copyright and the National Judicial Council; meetings with the presidents of the regional courts and courts of appeal; a delegation from the Japanese House of Representatives; a call to fill the Hungarian judge post at the European Court of Human Rights; the Family Bankruptcy Protection Service; and a meeting on 1 September 2026 with the leadership of the Budapest Regional Court.
Audit result: 54 – Grade D (Poor) · Structure A · Depth B · Schema F

Strongest findings:

• Structural Integrity is 100, and the alt-text check passes.
• The page names a broad set of institutions: the National Intellectual Property Office, the Hungarian industrial property association, the National Office for the Judiciary and its president, and the European Court of Human Rights.
• Entity Graph Stability is 70 (Moderate). Claude scores 80 (~46%) and ChatGPT 77 (~50%).
• 19 of 26 queries are flagged eligible, including the consultation with the National Judicial Council and the Family Bankruptcy Protection Service.

Weaknesses:

• No meta description, and Schema & Metadata is 20 (F).
• Freshness is 0, although the page names a specific meeting date (1 September 2026).
• Seven of 26 queries are full-sentence headlines tagged Digital Analytics at 45–49% eligibility, including the EPPO accession items.
• Entity labels are truncated: “Orosz Fruzsina”, “Senyei Gy”, “Hivatal Eln”, “Az Igazs”.
• The institutions named on the page (EPPO, the Judicial Council, the courts) have no schema, and neither does the minister. Gemini scores 37 (~21%) and Perplexity 45 (~46%).

What it means for AI systems: A justice ministry page is read as a source on institutions, accession processes and court leadership. It tells an engine which bodies are involved and nothing about who authored the page, when or on what authority.

Ministry of Education and Children’s Affairs

Website audited: https://kormany.hu/en/government/oktatasi-es-gyermekugyi-miniszterium
Role: The Oktatási és Gyermekügyi Minisztérium. The detected content covers the minister, Lannert Judit; the “Children’s and Students’ Voice” project day; open registration for the ministry’s public consultations; the leadership of the Klebelsberg Központ; the nursery needs survey (deadline October 22, extended); vocational students joining the National Student Parliament; the 2027 round of the Bursa Hungarica scholarship; the new school year; the reorganisation of higher education institutions and a new university model; applications for school district director positions; a consultation on ministerial awards; and a regional vocational forum series.
Audit result: 60 – Grade C (Fair), the highest in the set · Structure A · Depth B · Schema F

Strongest findings:

• It has the highest AI Retrieval Index of the five (60).
• It is the only page with any date signal. A date is detected through text or meta tags, content age passes the 180-day and 12-month checks, and update-frequency signals are present. No decay alert is raised.
• Gemini scores 45 (~22%) and Perplexity 53 (~48%), both the highest in the set. Claude scores 78 (~46%).
• 22 of 27 queries are flagged eligible, drawn from concrete announcements: the nursery survey, Bursa Hungarica 2027, the vocational student parliament seats and the higher education reorganisation.

Weaknesses:

• The dates are not in JSON-LD. The audit lists “Upgrade Date Signals to JSON-LD” and “Add JSON-LD Date Signals”. The JSON-LD dateModified, the explicit modification date, the publish and modified pair and the content-updated check all fail.
• No meta description, and Schema & Metadata is 20 (F). The only JSON-LD is a WebPage block.
• Balanced heading density fails (95).
• Entity Graph Stability is 67, with Co-occurrence at 12/15.
• The minister has a truncated biography query (“Lannert Judit – oktatási és gyermekügyi miniszter, Oktatáskutató, szociológus és szakpolit”) and no Person schema.
• Two announcements are detected twice, in headline and “What is” forms (the open registration notice and the project day).

What it means for AI systems: This page shows what a date signal does and does not buy. It is the best page in the set, and a date outside JSON-LD earns only a Grade D on freshness. The deadlines on the page, such as the October 22 survey, are the kind of fact an engine needs to read as current.

What This Means

Hungary’s public knowledge layer is structured and readable, and undeclared. All five pages parse cleanly: one H1, a clean hierarchy and lists. None names a publisher, an author or a date in structured form. KG Anchoring is 0 at all five, Organization and Person schema is absent everywhere, and no E-E-A-T signal registers.

Hungary’s AI visibility is set by the platform, not by the ministry. All five pages sit on kormany.hu and return the same markup pattern: one WebPage block, no Organization, no Person, no Article. The index spread is 6 points (54–60), against 24 points in Greece (37–61). Entity Clarity is 55 and Copilot’s score is 66 at all five. Every ministry will read the same way to AI until the platform changes. The changes in this profile apply once and reach all five pages.

Freshness is the national failure. The average is 9.0, the lowest dimension in the set. Four of five pages, including the government roster and three ministries, carry no date a machine can read. These pages hold current announcements, meeting dates and deadlines: the November 30 licence plate deadline, the October 22 nursery survey, the 2027 scholarship round.

Structure won Hungary the lead over Greece, and it only goes so far. Hungary scores 6.4 points above Greece because it fixed what Greece lacks, a single H1 and a logical hierarchy. That lifted Claude, ChatGPT and Copilot by 10 to 18 points. It left Perplexity and Gemini where they were, and Gemini’s average of 39.6 matches Greece exactly. The two engines that need structured data, authorship and dates have not been given any.

The entity layer is empty. The government page names ministers, and the ministry pages name their ministers, institutions and programmes. None of them is declared as a Person or an Organization. A machine reads these as strings, truncated in places, with no link to Wikipedia, Wikidata or any other identifier.

The fixes are metadata, and they are the same fixes for all five.

• Add datePublished and dateModified to JSON-LD at all five, including the Education page, where dates exist outside JSON-LD.
• Add Organization schema for each ministry, with sameAs links to Wikipedia or Wikidata.
• Add Person schema for each minister and Article schema for the news items.
• Write a meta description for the four pages without one.
• Add alt text to the 19 images on the Government page, and tables or structured data on all five.
• Fix balanced heading density on the Government and Education pages.

Hungary and Greece fail on different dimensions. Greece had extractable pages with no H1, and Hungary has well-built pages with no declared layer. The Hungarian pages are ready for AI to read, and AI cannot yet verify what it reads.

Scope and Method

What was audited. Five Hungarian government pages, one page each, using the AI Visibility Inspector v1.9.3 (Srna SEO Frameworks, srnaseo.com). All five audits were generated on 12 October 2026. The five pages are the government overview page and four ministry pages, all on the kormany.hu platform.

What the research measures. Each report gives an AI Retrieval Index and grade, Structure, Depth and Schema sub-grades, and five scored dimensions: Structural Integrity, Data Extractability, Entity Clarity, Schema & Metadata, and Freshness & Decay. It adds scores for five engines (Perplexity, OpenAI / ChatGPT, Claude, Google Gemini, Microsoft Copilot) with estimated citation percentages, a Schema Intelligence section (validation, entity connectivity, KG anchoring, E-E-A-T density, semantic richness), query eligibility and an entity graph stability score.

Boundaries.

• One page per institution. Subpages, PDFs and other domains are not covered.
• Findings describe each page on the audit date.
• Averages are simple means of the five institutional scores. Greek and German comparison figures are those published in the Greece profile.
• The five pages share one domain and one markup pattern, so they are not five independent sites. The profile reports each page separately and reads them as one platform where the evidence points that way.
• All five URLs sit under the /en/ path, and the extracted text and queries are in Hungarian.
• The reports produce two freshness blocks per page. The scores matched within each institution.
• The Government page scores 80 on Data Extractability with its alt-text check failing, while the other four score 75 with it passing. The Inspector does not reconcile the two.
• Entity and query detection is automated and produces artefacts. Hungarian phrases are wrapped in an English “What is” template, interface text is read as queries, and entity labels are cut at the first accented character. Entity labels and queries are quoted as the audit displays them.
• Every report lists the same ten secondary content gaps (for example Artificial Intelligence, Large Language Model, Dark Funnel). The list is generic and is not interpreted here.
• Citation figures are the Inspector’s estimates, not observed citations in live engine answers.
• People named in the institution profiles appear because they were detected in the audited pages’ queries. English renderings of institution names and Hungarian text are editorial translations.

How Does Your Institution Read to AI?

The five Hungarian pages above were read with the same tool available to any organization. The AI Visibility Inspector produced each report in this profile.

Run your own homepage and compare it with the five profiles above: the H1 count, the schema blocks, the dates, the Person and Organization declarations and the five engine scores. If the pattern matches, the cost is the same. The AI Visibility Cost of Inaction Calculator lets you enter your own figures and see the exposure for your organization.

For a deeper reading, the Knowledge Exposure Audit and the Agent Exposure Audit extend the inspection beyond a single page. For fixes, see the relevant frameworks.

Calculate Your Potential Exposure

Reading poorly to AI has a cost. The AI Visibility Cost of Inaction Calculator lets you enter your own figures and see the exposure for your organization.