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

Why Norway, and Why Now

Norway’s five audited pages average 58.0 on the AI Retrieval Index, the highest of the five countries compared in this series so far. Ireland, the previous profile, averaged 51.8; Hungary 55.8, Greece 49.4 and Germany 51.8.

Five institutions speak for the Norwegian state in different ways: the government (regjeringen.no), the armed forces (forsvaret.no), Statistics Norway (ssb.no), the Data Protection Authority (datatilsynet.no) and the Norwegian Mapping Authority (kartverket.no). When someone asks an AI assistant what is in the national budget, how large a population is, how a privacy fine was decided or where a property boundary runs, these are the pages the answer should come from.

The audits measure three things about each page: whether an AI system can read it, whether it can tell who published it, and whether it can judge that the content is current. All five audits were generated on 12 October 2026.

The main finding: Norwegian public pages are built cleanly and declare almost nothing about themselves. Structure averages 92.0 and extractability 71.4. Schema averages 31.0. None of the five declares a Person or Organization in JSON-LD, none links to Wikipedia or Wikidata, and none registers a single expertise, experience, authority or trust signal.

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, Ireland, Norway.
Read also Norway Flagship companies research.

The Profile in One View

The spread between the best and worst page is 27 points. Three pages grade C and two grade D. No page reaches B.

Norway leads on the headline number, but the lead is built from structure and freshness, not from schema. Against Germany, which averages 51.8, Norway has 18.0 points more structure and 10.4 more freshness, 3.0 fewer on Entity Clarity and the same schema score (31.0). Against Hungary (55.8), Norway is 6.0 points behind on structure and 3.0 behind on Entity Clarity, and 22.8 points ahead on freshness.

Other indicators across the five pages:

• Entity graph stability averages 44.0. Scores run from 38 (Forsvaret, “unstable”) to 49 (Datatilsynet and Kartverket, “fragmented”). The schema sub-score is 0 of 20 on all five.
• Entity connectivity is 0 of 100 on four pages. Only SSB scores above zero, at 30.
• KG anchoring and E-E-A-T density are 0 of 100 on all five pages.
• Three of five pages carry no JSON-LD at all. The two that do carry three blocks between them, and one of those blocks is valid.

How the Five Engines Read the Norwegian Pages

Each audit scores the page for five AI engines and estimates how likely each is to cite it. Averaged across the five Norwegian pages:

Per institution, as score (estimated citation):

Claude and ChatGPT read Norway best, because both reward the thing Norwegian sites do well: clean, hierarchical pages. Claude scores 75.0 on average and ChatGPT 72.0. Four of five pages score 75 or higher with Claude. The exception is Datatilsynet at 60, where two H1 tags create what the audit calls an ambiguous document root, and its estimated citation drops to ~24% against 46–51% for the other four.

Gemini is the weakest reader at 45.4, with an estimated citation rate near 20%. Gemini asks for Article and Person JSON-LD, a Person and Organization pairing, and complete image alt text. All five pages miss the pairing. Alt-text coverage is partial on four pages: 77% on Forsvaret, 75% on Kartverket, 63% on Datatilsynet and 29% on SSB. Datatilsynet’s Gemini score of 27 is the lowest single engine score in the set.

Perplexity depends on how much schema and canonical signal the page carries, and the scores show it. SSB reaches ~62% estimated citation and is the only page where the audit reports core citation signals detected. Regjeringen follows at ~54%. The other three sit at ~30–35%. Four of the five pages have no canonical URL, which the audit flags as a citation deduplication risk.

Copilot rewards the one Norwegian page that links its publisher. SSB scores 81 with a ~57% estimate, against 42–60 for the other four. SSB’s Article block names a publisher, which is the only entity link in the set. It still lacks a valid author, and the audit asks for both Person and Organization schema.

ChatGPT’s findings are about depth and naming. Three pages fall under the 600-word threshold the audit uses: Datatilsynet (479 words), Forsvaret (527) and SSB (576). Four of five carry the note that entity signals are weak and that key topics should be named and described explicitly. Only SSB does not.

What the Five Audits Reveal

Structure is the national strength (92.0, range 70–100). Forsvaret and Kartverket score 100, Regjeringen and SSB 95. Each has one H1, H2 sub-sections and logical heading nesting. Datatilsynet scores 70 because it has two H1 tags, the only item the audit marks critical in the set.

Extractability is good but uneven (71.4, range 64–85). Regjeringen scores 85 and Kartverket 78. Forsvaret scores 66, SSB 64 and Datatilsynet 64. Four pages have images without alt text: 9 on Forsvaret, 10 on SSB, 6 on Datatilsynet and 2 on Kartverket.

Entity clarity is middling (52.0, range 40–75). SSB leads at 75 and is the only page with an Article block. Regjeringen scores 55, Forsvaret and Kartverket 45, Datatilsynet 40. All five pages tokenise their own name correctly. What they lack is the schema layer that would tell a machine what kind of entity the name belongs to.

Schema and metadata is the weakest of the build dimensions (31.0, range 15–65). SSB scores 65. The other four score 25, 25, 25 and 15. Regjeringen has two JSON-LD blocks, a WebSite block that fails validation for a missing name and a valid WebPage block. SSB has one Article block that fails for a missing author, with 4 of 8 Article fields present. Forsvaret, Datatilsynet and Kartverket have no JSON-LD. Four of five pages have no canonical URL. SSB is the exception: the audit does not flag one.

Freshness is where the five pages split (31.8, range 0–65).

• Forsvaret scores 0 and Datatilsynet 3. Neither page has a machine-readable date.
• SSB scores 27. The audit reads its content as 2,574 days (7.1 years) old with no update signal.
• Regjeringen (64) and Kartverket (65) show dates, but not in JSON-LD.

No page has dateModified in JSON-LD. The two best freshness scores are earned without the signal the audit recommends most often.

The trust layer is empty. Entity connectivity is 0 on four pages and 30 on SSB. KG anchoring is 0 everywhere because no page has a sameAs link to Wikipedia or Wikidata. E-E-A-T density is 0 everywhere, with zero expertise, experience, authoritativeness and trustworthiness signals detected on each page. Semantic richness is 50 on Regjeringen and 31 on the other four.

The entity graph recognises the name and nothing around it. Every page scores 20 of 20 on primary entity and 15 of 15 on clarity. Secondary entities score 3–9 of 20, co-occurrence 0–2 of 15, and schema 0 of 20. Totals run from 38 to 49, and every page is labelled either “fragmented” or “unstable”.

Query eligibility is flat at about 50%, with one exception. The audits generated 132 queries across the five pages. Average eligibility is 50.3–50.7% on four pages and 67.5% on SSB, whose queries range from 65% to 72%. SSB’s headline phrase, “Offisiell statistikk siden 1876”, is the highest-scoring query in the set at 72%.

Automated detection picks up interface noise. Queries and entities in these reports include cookie-banner text on Forsvaret and SSB, an “Aktiver Javascript” entity on Regjeringen, CMS layout labels on SSB (“Forside banner, Rad A”, “Nøkkeltall, Rad B”), a “Conversion Rate” topic on Datatilsynet and a “Google Analytics Iframe” entity on Kartverket. Several headline queries also lose their subject, such as “Varsler ny nasjonal strategi for kunstig intelligens” on Regjeringen. These are artefacts of how the pages are marked up, and the Boundaries section says how far to trust them.

Institution Profiles

Statistics Norway (ssb.no)

SSB is Norway’s official statistics agency, and the audit reads its homepage as being about “Statistisk sentralbyrå”. It scores 68, grade C – Fair, the highest in the set.

Strongest findings. It is the only page with an Article block, with headline, datePublished, publisher and description present (4 of 8 fields), and the only page with any entity connectivity (30 of 100). Its estimated citation is the highest of any page on Perplexity (~62%) and Copilot (~57%). Its 39 queries average 67.5% eligibility against about 50% elsewhere.

Weaknesses. The Article block fails validation because it has no author. The audit reads the content as 2,574 days (7.1 years) old with no update signal and flags it as severe freshness decay. Ten images lack alt text, leaving coverage at 29%, the lowest in the set. There is no Person or Organization schema and no sameAs link.

What it means for AI systems. SSB is the closest Norwegian page to being citable as it stands. An engine can see who published it, but not who stands behind it and not whether it is current. Adding an author, a dateModified and an Organization block with sameAs links would close the three gaps the audit names, without touching the page’s content.

Government of Norway (regjeringen.no)

The audited page is the Norwegian-language front page (regjeringen.no/no/id4/), which the audit reads as being about “Forsiden”. It scores 66, grade C – Fair.

Strongest findings. It has the best extractability in the set (85) and the best semantic richness (50 of 100). It carries two JSON-LD blocks, and the WebPage block is the only valid block across all five audits. Dates are present and the content is fresh. The ChatGPT note reads its depth and structure as optimised for semantic retrieval.

Weaknesses. The WebSite block fails validation because it has no name. There is no canonical URL, no Article, Person or Organization schema, and no sameAs links. The dates sit outside JSON-LD. The entity detector finds “Regjeringen.no” and an “Aktiver Javascript” notice, and no topic clusters. Its 40 queries are mostly headline fragments and minister names, such as “Stubb og Støre diskuterte KI” and “Varsler ny nasjonal strategi for kunstig intelligens”. The second has lost its subject.

What it means for AI systems. The page carries news about a national AI strategy and a Støre–Stubb discussion on AI, and an engine reading it finds no topic cluster for either. The facts are on the page. The machine-readable layer that would attach them to the Norwegian government as a named publisher is missing.

Norwegian Mapping Authority (kartverket.no)

The audit reads the homepage as being about “Kartverket.no”. It scores 64, grade C – Fair.

Strongest findings. Structure is perfect at 100 and freshness is the best in the set at 65, with dates detected in text or meta tags. Only two images lack alt text. Its ChatGPT score is 75, with an estimated citation of ~50%, second only to Regjeringen’s ~51%.

Weaknesses. There is no JSON-LD at all, so no Article, Person or Organization block and no canonical URL. The audit asks for the dates to be moved into datePublished and dateModified fields. The entity detector finds “Kartverket.no”, a generic “Product / Service” entity and a “Google Analytics Iframe”, and no topic clusters. Gemini scores the page 49 with a ~18% estimate.

What it means for AI systems. Kartverket’s homepage lists concrete news items, among them funding of more than 126 million kroner for a property initiative, climate profiles for 357 municipalities and better 3D in Norgeibilder. In the audit they surface as headline fragments without a named subject. The page has what a mapping authority needs most, current content with a clear structure, but it carries no Organization markup that names the publisher.

Norwegian Armed Forces (forsvaret.no)

The audit reads the homepage as being about “Forsvaret”. It scores 51, grade D – Poor.

Strongest findings. Structure is perfect at 100: one H1, H2 sub-sections and clean nesting. Claude scores the page 75, with an estimated ~46% citation.

Weaknesses. The audit finds no date signal of any kind and marks the page for structural decay, which produces the freshness score of 0. There is no JSON-LD, no canonical URL and no Person or Organization schema. Nine images lack alt text, so coverage is 77%. The page has 527 words, which the audit calls moderate depth. The entity graph scores 38 (“unstable”), the lowest in the set, with a single detected entity and no co-occurrence signal. Gemini scores it 36 (~15%) and Perplexity 44 (~32%).

What it means for AI systems. An engine that reads the page cannot tell whether it describes the armed forces as they are today or as they were two years ago. The page is easy to parse and impossible to date.

Norwegian Data Protection Authority (datatilsynet.no)

The audit reads the homepage as being about “Datatilsynet”. It scores 41, grade D – Poor, the lowest in the set.

Strongest findings. Its entity graph (49) ties for the best in the set, with a secondary-entity score of 9 of 20 and a query-alignment score of 3 of 10, both shared with Kartverket. The audit rates its heading hierarchy as acceptable apart from the double H1.

Weaknesses. Two H1 tags are the audit’s only critical finding in the set. Structure scores 70 (grade C), the only page below A. There is no JSON-LD, no canonical URL and no machine-readable date. Six images lack alt text, so coverage is 63%. The page has 479 words, the lowest of the three audits that report a word count. Schema & Metadata scores 15, the lowest in the set. Claude scores it 60 (~24%), Gemini 27 (~14%) and Copilot 42 (~26%), the lowest in every engine column.

What it means for AI systems. The data protection regulator publishes the least machine-readable homepage of the five. Two queries, the welcome heading and a headline about a 92 million fee for a Microsoft company, are tagged ANALYTICS rather than DATATILSYNET, and the only topic the detector finds is “Conversion Rate”. That suggests tracking markup is being read as page content. Merging the H1s and adding an Organization block are small changes that address the audit’s critical finding and its most repeated one.

What This Means

Norwegian public pages are easy for AI to read and hard for AI to identify. Structure scores 92.0 on average. Entity connectivity, KG anchoring and E-E-A-T density are 0 on at least four of five pages each. The pages tell an engine what they say. They do not tell it who is saying it.

One Organization block with sameAs links is the highest-leverage fix. Claude, Gemini and Copilot each carry a note about missing Person or Organization schema on every one of the five pages. The same block addresses the entity connectivity and KG anchoring scores that are 0 on all or nearly all of them.

Freshness is the cheapest lift, because on three of five pages the content already exists. Regjeringen, Kartverket and SSB all show dates or publication data. Moving them into JSON-LD with a dateModified is a small change. Forsvaret and Datatilsynet first need a visible date to mark up.

Most of the gaps the audits name are specific items, not rebuilds. Datatilsynet’s double H1, Forsvaret’s missing dates and SSB’s missing author are each a single line on the Inspector’s action list.

Scope and Method

What was audited. Five Norwegian public institutions, one page each, using the AI Visibility Inspector and Srna SEO Frameworks. All five audits were generated on 12 October 2026.

• regjeringen.no/no/id4/, the Norwegian-language front page of the government
• forsvaret.no, the Norwegian Armed Forces
• ssb.no, Statistics Norway
• datatilsynet.no, the Norwegian Data Protection Authority
• kartverket.no, the Norwegian Mapping Authority

These five were chosen because they are Norway’s public doors for government, defence, statistics, data protection and mapping.

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, English and Sámi versions and other domains are not covered.
• Findings describe each page on the audit date. The headlines quoted are those on the page on 12 October 2026.
• Averages are simple means of the five institutional scores. Ireland, Hungary, Greece and Germany comparison figures are those published in the earlier profiles.
• Each report produces two freshness blocks per page. The scores matched within each institution.
• The audit’s scoring is not fully reconciled. Regjeringen and SSB both score 62 on Perplexity, with estimated citations of ~54% and ~62%, and the Inspector does not explain the difference.
• SSB’s content age (7.1 years, 2,574 days) is the Inspector’s reading, and the report does not say which date it read.
• The Inspector reads the page as delivered. The “Aktiver Javascript” entity on Regjeringen suggests at least part of that page was read without running scripts. If any of these sites adds JSON-LD with JavaScript after load, the audits cannot show it.
• Entity and query detection is automated and produces artefacts. Interface text is read as queries and entities, Norwegian phrases are wrapped in an English “What is” template, some queries are cut off mid-sentence, and “Product / Service” is read as a commercial entity. Queries and entity labels are quoted as the audit displays them.
• Every report lists the same ten secondary content gaps, such as Large Language Model and 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.
• Headlines, figures and named items in the institution profiles are those detected on the audited pages. They are reported as page content and are not independently verified.

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.