In This Article
Five Czech government homepages share a domain pattern and a shared blind spot. AI can read what each page calls itself, but the audited pages provide surprisingly little structured evidence about who is speaking, which institution is responsible, how current the information is, and how the institutions connect to Czechia’s wider AI governance system.
In This Article
- National average: 50.4, Grade D. This is the second-highest of the five national profiles so far, behind Cyprus (69.4). The range is 39 to 64. No site reaches Grade B.
- The Digital and Information Agency (DIA) leads at 64, and the national portal sits at 39. The gap is not content quality. It is structure. DIA has a single H1, a canonical URL and an Open Graph title. The portal has none of these.
- Still zero JSON-LD on all five. Entity connectivity, knowledge-graph anchoring and E-E-A-T density are 0/100 everywhere. No Organization, Person or Article schema exists on any homepage.
- Four of five carry competing H1 tags. The portal has no H1 at all. DIA is the only site with a clean structure score of 100. The others range from 55 to 100.
- Freshness is absent or unverifiable. Four sites have no machine-readable date signal. DIA has dates detected via meta tags but not in JSON-LD. On the four sites without dates, engines treat the content as stale regardless of editorial reality.
- Czechia has a Government Commissioner for AI, based at MPO. The position was created by government resolution in May 2025, and Lukáš Kačena was appointed in January 2026 . MPO coordinates the National AI Strategy, and the Commissioner coordinates AI development across government . None of this appears in structured data on any audited homepage.
The profile in one view
Scores are the Inspector’s dimension scores (0 to 100) for each homepage.
| Area | DIA | vlada.gov.cz | mv.gov.cz | mpo.gov.cz | portal.gov.cz |
|---|---|---|---|---|---|
| Content architecture | 100 | 100 | 100 | 95 | 60 |
| Data extractability | 75 | 85 | 65 | 90 | 75 |
| Entity clarity | 45 | 45 | 45 | 40 | 40 |
| Schema and metadata | 35 | 25 | 20 | 0 | 10 |
| Freshness signals | 53 | 0 | 0 | 0 | 0 |
| Trust and authorship | 0 | 0 | 0 | 0 | 0 |
| Overall index | 64 (C) | 55 (C) | 50 (D) | 49 (D) | 39 (D) |
Shading: 75 and above strong, 50 to 74 moderate, 20 to 49 weak, below 20 absent. The sites are dia.gov.cz, vlada.gov.cz, mv.gov.cz, mpo.gov.cz and portal.gov.cz.
Czechia against Austria, Belgium, Croatia and Cyprus
Averages of the five homepages in each country.
| Measure | Austria | Belgium | Croatia | Cyprus | Czechia |
|---|---|---|---|---|---|
| Overall index | 49.0 | 49.8 | 40.4 | 69.4 | 50.4 |
| Range (low to high) | 44 to 55 | 35 to 77 | 38 to 46 | 66 to 81 | 39 to 64 |
| Sites above Grade D | 1 | 1 | 0 | 5 | 0 |
| Sites with any JSON-LD | 0 | 1 | 0 | 0 | 0 |
| Sites with competing H1s | 1 | 1 | 4 | 4 | 4 |
| Content architecture | 94.0 | 94.0 | 76.0 | 64.0 | 91.0 |
| Data extractability | 69.8 | 52.6 | 55.6 | 53.2 | 79.0 |
| Entity clarity | 45.0 | 47.0 | 44.0 | 100.0 | 43.0 |
| Schema and metadata | 15.0 | 21.0 | 12.0 | 85.0 | 18.0 |
| Freshness signals | 3.2 | 21.0 | 0.8 | 45.0 | 10.6 |
| Trust and authorship | 2.6 | 3.4 | 0.0 | 0.0 | 0.0 |
Czechia is the mirror image of Cyprus. Cyprus had strong metadata (85.0) and weak architecture (64.0). Czechia has strong architecture (91.0) and weak metadata (18.0). Both profiles share the same absence: zero JSON-LD, zero trust signals, and no machine-readable identity for the institutions that publish national AI policy.
The score distribution tells the real story. Cyprus had no site below Grade C. Czechia has no site above Grade C. The best Czech homepage (DIA, 64) would rank below all five sites in the Cypriot set. The weakest Czech homepage (portal.gov.cz, 39) would rank below Croatia’s average.
Where Czechia needs to improve
Ranked by how many institutions are affected and how much one fix would unlock.
1. Structured identity (all five)
No Organization, Person or Article schema exists on any site. The Government Commissioner for AI is based at the Ministry of Industry and Trade, and the position is defined by government resolution . The Commissioner coordinates the National AI Strategy Action Plan, Czech adaptation to the EU AI Act, and the national supervisory and support bodies . The Commissioner is Lukáš Kačena, appointed in January 2026 . None of this is marked up as Person schema on any homepage.
MPO is the gestor of AI Act implementation and coordinates the National AI Strategy . DIA publishes an AI Act glossary and an AI use catalogue for public administration . The content is there. The structured identity is not.
The audited homepages provide no structured entity relationship connecting the institutions, their officials and their AI policy responsibilities. In a domain where authority matters – AI regulation, national strategy, data governance – this is the single highest-leverage gap.
2. Dates that do not exist in machine-readable form (four of five)
vlada.gov.cz, mv.gov.cz, mpo.gov.cz and portal.gov.cz have no JSON-LD dateModified and no machine-readable date signal the Inspector can credit. DIA has dates detected via meta tags but not in JSON-LD, which holds its freshness score at 53 against a possible 100.
The content is not necessarily stale. MPO published an AI Committee press release in September 2026. MV published a project update in January 2026. But without dateModified in JSON-LD, engines that weigh recency – Perplexity, Gemini – have no way to know this. The Inspector therefore treats the homepage’s freshness as unverifiable, even though recent activity exists elsewhere on the site.
3. Competing H1 tags (four of five)
The portal has no H1 at all, which is worse than competing H1s. vlada.gov.cz, mv.gov.cz and mpo.gov.cz carry multiple H1 tags, holding their structure scores at 100 only because the Inspector’s balance metric credits them. DIA’s single H1 is the only clean anchor in the set.
Claude is the engine most sensitive to heading disorder. Its scores show the cost: 80 on vlada.gov.cz against 77 on DIA, despite DIA having the lower overall index. The portal’s missing H1 is a structural decay alert, and Claude reports it as a document anchor completely absent.
4. The national portal is the weakest site (portal.gov.cz, 39)
The front door of Czech eGovernment scores Grade D. It has no H1, no JSON-LD, no canonical URL the Inspector can credit, a 10/100 schema score and a 42% eligibility rate across 33 candidate queries. Three of its queries are interface text: “Můžeme si u vás uložit cookies?” and “Nastavení souhlasů s cookies” are cookie-banner sentences, not citizen questions.
This is the page that should be the machine-readable root of Czech government. Czechia’s portal scores 39 and is the weakest link in this five-site sample. Czechia’s portal scores 39 and serves as the weakest link. A WebSite entity and a GovernmentOrganization entity on portal.gov.cz, linked to the relevant official government entities, could establish a stronger machine-readable root for Czech eGovernment. Without it, each site stands alone.
5. The AI governance story is invisible to AI (all five)
Czechia has a substantial and active AI policy apparatus. The Government Commissioner for AI was established by government resolution in May 2025, and Lukáš Kačena was appointed to the role in January 2026. The National AI Strategy 2030 was approved in July 2024. The AI Committee, chaired by the Government Commissioner, coordinates implementation of the strategy, while MPO leads major elements of AI policy and AI Act implementation. DIA operates an AI use catalogue and AI glossary, and ČTÚ launched an AI Supervision Unit in September 2026.
The underlying governance work is visible in the content, but the audited homepages do not expose the key institutions, people, mandates and policy relationships through structured data. A machine reading mpo.gov.cz sees “Ministerstvo průmyslu a obchodu” as a 40-clarity entity and “Artificial Intelligence” as the only detected concept. It does not see the Commissioner, the strategy, the committee or the AI Act implementation as connected entities with authorship and dates. The work is real. The signal is absent.
Institution profiles
Digitální a informační agentura – Grade C – 64/100
The best-structured site in the set, still invisible to engines
Strengths: The only homepage with a single H1 and a structure score of 100. The strongest Copilot reading (58) and the highest ChatGPT citation estimate among Czech sites. Dates are detected via meta tags, giving it the only non-zero freshness score (53). DIA publishes AI governance infrastructure: an AI Act glossary and an AI use catalogue for public sector projects . It also hosts the digital and information agency’s own work on data management and eGovernment services.
What AI sees: “DIA” as an organization with 68% confidence, and nothing else. The AI Act glossary is a page with no Article schema. The AI use catalogue is a list with no Dataset markup. Entity clarity sits at 45. The page is about digital government and AI governance, but no structured data says so.
Missed opportunity: DIA is the institution that will make Czech government data machine-readable through its catalogue and glossary work. Its own homepage is the proof of concept that does not exist. An Organization record with sameAs links to the European Commission’s AI Act pages, a Dataset markup for the AI use catalogue, and a Legislation schema for the AI Act glossary would let engines cite DIA as the authoritative source for Czech AI governance. The content is already there. The markup is not.
vlada.gov.cz – Grade C – 55/100
The page that publishes the Commissioner’s mandate, with no Person schema
Strengths: A structure score of 100 and a data extractability score of 85. The page publishes the Government Commissioner for AI’s mandate and links to government resolutions . Claude scores it at 80, the highest of any Czech site on that engine. The content is authoritative by any human reading.
What AI sees: “Úvod” as the primary topic – the Czech word for “Introduction” – because no H1 anchors the page to a real entity. The AI Commissioner is a page section, not a Person entity. The government resolutions that created the position are referenced text, not Legislation schema with datePublished and dateModified. Entity clarity sits at 45, same as DIA and mv.gov.cz, despite hosting the most nationally significant AI governance content in the set.
Missed opportunity: vlada.gov.cz is the canonical source for the Commissioner’s mandate and the government resolutions that define Czech AI governance . A Person schema for Lukáš Kačena with role (Government Commissioner for AI), affiliation (Ministry of Industry and Trade), and sameAs links to official government pages would make the human authority verifiable. Structured metadata for the relevant government resolutions, including dates, issuing authority and relationships to the Commissioner role, would make the policy framework easier for machines to connect and cite. Without these, engines see a page about “Introduction.”
mv.gov.cz – Grade D – 50/100
The ministry that publishes internal AI adoption, with the weakest entity layer
Strengths: The most detailed public account of AI adoption inside a Czech ministry. mv.gov.cz documents AI-assisted document routing and a 50 million CZK project for AI integration in document management. It publishes a public administration magazine that covers AI as a priority. The content is specific, named and dated in human-readable form.
What AI sees: “Ministerstvo vnitra” as an entity with 68% confidence, plus five co-occurrence fragments: “Miroslava Tomana,” “Martina Legnera,” “Ministr Metnar,” “Andrejem Babi” and “Design System.” No Person schema. No Organization schema. No Article schema for the project documentation. Entity clarity sits at 45. The AI governance work is visible as text and invisible as structure.
Missed opportunity: The Ministry of Interior is the institution that runs the document digitization project and publishes the public administration magazine. A Project schema for the digitization initiative with funding source, timeline and status would make the work citable. A Person schema for the named officials would connect the policy to the people. The content already names them. The markup does not.
mpo.gov.cz – Grade D – 49/100
The ministry that owns AI policy, with the lowest schema score
Strengths: The highest data extractability in the set (90) and the strongest ChatGPT and Claude readings (77 each). The Ministry of Industry and Trade is the gestor of AI Act implementation, coordinates the National AI Strategy, and houses the Government Commissioner for AI . It has a structured, well-organized page with a 95/100 structure score.
What AI sees: “Ministerstvo průmyslu a obchodu” as the page topic, “Artificial Intelligence” as a single detected entity with 73% confidence, and a schema score of 0. The AI Committee press release, the National AI Strategy and the AI Act implementation documents are all readable text with no Article, GovernmentService or Legislation schema. Every key policy document is a name without machine-readable metadata.
Missed opportunity: mpo.gov.cz is the institution that coordinates Czech AI policy and employs the Commissioner. Its press releases cover AI Committee meetings, AI Gigafactories, and the AI Act implementation timeline. None of this is marked up. A NewsArticle schema for each press release with datePublished and author, plus a GovernmentService schema for the AI sandbox and a Legislation schema for the draft AI law, would turn the ministry’s policy work into citable, verifiable entities. The 0/100 schema score is the cost of not doing this.
portal.gov.cz – Grade D – 39/100
The front door of Czech eGovernment, with the weakest structure
Strengths: The page is the national access point for citizen services. It has a 75/100 data extractability score and a 60/100 content architecture score. Its candidate queries describe real citizen tasks: benefits, ID cards, driving records, criminal records, digital identity.
What AI sees: “Gov.cz” as the page topic with no H1 anchor. No JSON-LD. No canonical URL credited. A 10/100 schema score. Three of its 33 candidate queries are cookie-banner text. Entity clarity is 40, the lowest in the set. The portal is the site most engines would cite for “how do I get a Czech ID card” and it has no structural foundation to be cited.
Missed opportunity: Cyprus’s portal scored 81, providing a useful contrast: its stronger machine-readable metadata shows what a government portal can achieve even when its content architecture is less consistent. Czechia’s portal scores 39 and needs the fix the ministries already have. A WebSite schema, a GovernmentOrganization record with sameAs links to the ministries, and a single H1 would raise the portal above Grade C. The queries are already there. The structure is not.
Scope and method
Each institution’s English-language homepage was audited once on 5 October 2026 with the AI Visibility Inspector (v1.9.3) and the Ivica Srncevic Framework. This is a homepage snapshot, not a full-site audit. All five sites share the gov.cz domain. The shared-template reading is an inference from identical scores on entity clarity, schema and metadata, and trust and authorship, and should be confirmed in the page source. The Inspector marks some entities with confidence scores and credits entity clarity in the 40s while its own schema check reports no JSON-LD. This profile treats JSON-LD as absent and reads the credit as coming from page-level metadata and entity tokenization. The “content gaps” list is identical on all five and was left out. Engine citation percentages are the tool’s estimates.
The national average of 50.4 is the arithmetic mean of the five overall index scores: 64, 55, 50, 49 and 39.
This research series also includes National AI profiles of Austria, Belgium, Croatia, Cyprus and Czechia.
Read also Czech Flagship Companies.
Calculate your potential exposure below
Inputs
Enter your own data where possible. Hover over labels or open “Methodology” for details on how each field is used.
From GA4, GSC, or your SEO platform. Use a 30–90 day average.
Session → lead / signup / purchase. Use decimal (e.g. 1.2 for 1.2%).
For B2B, use average deal value or LTV. Currency is symbolic here.
Used to show profit‑based cost of inaction. Leave at 100 to ignore.
Share of organic traffic from queries/pages likely to be influenced by AI answers and retrieval systems.
Based on observed traffic loss patterns in AI search and structural decay cases.
Expected organic growth if you do nothing special about AI visibility.
Plausible uplift over the chosen horizon based on case studies.
How these numbers are calculated (methodology)
Let:
- O₀ = monthly organic sessions
- CVRₒ = organic conversion rate (as a decimal)
- ARPC = average revenue per conversion
- GM = gross margin (as a decimal)
- E = AI exposure share (as a decimal)
- D = displacement rate (as a decimal)
- T = time horizon in months
Baseline monthly revenue:
- Revenue: \(R_0 = O_0 imes CVR_o imes ARPC\)
- Profit: \(R_{0,profit} = R_0 imes GM\)
Monthly revenue at risk (for a given D):
- \(R_{risk} = R_0 imes E imes D\)
- Profit version: \(R_{risk,profit} = R_{risk} imes GM\)
The tool computes R_risk for three scenarios:
- Conservative: D = 0.15
- Base: D = 0.30 (or your selected scenario)
- Aggressive: D = 0.50
Cost of Inaction over T months is approximated as:
- \(CoI(T) = R_{risk} imes T\)
- Shown as a range from conservative to aggressive.
The growth inputs (g₀, g₁) are currently used for narrative context and can be incorporated into a more advanced version that models opportunity cost explicitly.
