AI Readiness Benchmark 2026: 447 Norwegian trade businesses
We crawled and scored the websites of 447 Norwegian plumbing and electrical businesses with our 33-test methodology. The headline finding: AI can read almost every website — and understands very few of them.
About the data
The population comes from the Brønnøysund Register Centre: Norwegian businesses in plumbing and electrical installation with five or more employees. Websites that were unreachable were excluded; 447 remained — 123 plumbing and 324 electrical businesses, crawled in July 2026.
Scores use methodology v0.2. The benchmark crawl runs the automated categories A–F (92 of the 100 points); category G — what AI assistants actually answer — is tested by hand per business, not at scale, so it contributes zero here for everyone equally and the maximum achievable score in this dataset is 92.
How the scores are distributed
Number of businesses per AI Readiness Score band (447 businesses, July 2026). No business scored below 30 or above 89.
Most of the market sits in the 60–79 band: a decent technical baseline with clear, fixable gaps. Only 19 businesses (4%) reach the AI-ready band — and 8 score below 40, effectively invisible to AI assistants.
| Industry | Businesses | Median | Middle half | Top score |
|---|---|---|---|---|
| Plumbers | 123 | 70 | 60–74 | 85 (ØKO AS) |
| Electricians | 324 | 67 | 61–72 | 84 (EL-CON AS) |
The problem is not access — it is understanding
Average share of points earned per category, across all 447 businesses:
Category G (live AI answers) is not measured in the bulk crawl — see "About the data" above.
The top of the list is the good news: Norwegian trade websites are open to AI. Robots.txt files almost never block AI crawlers (97% of category-A points earned), and the content is overwhelmingly readable without JavaScript (91%). Hardly anyone is locked out.
The bottom is the story. Structured data — the machine-readable layer that tells an AI assistant who the business is, what it does and where — earns just 31% of its points. Content structure (question-shaped headings, FAQ sections that match how customers actually ask) earns 47%. In other words: the door is open, but the room is unlabelled.
The checks businesses fail most
Share of the 447 businesses passing each check:
The most striking number: not one of the 447 businesses ships FAQ structured data — the single cheapest way to hand AI assistants ready-made answers about services, coverage area and pricing. Three quarters do not publish their organisation number in page content, leaving AI to guess which legal entity the website belongs to.
What this means for a trade business
The bar is low. With a median of 68 and only 4% of the market AI-ready, the businesses that fix the semantic layer — structured data, question-shaped content, unambiguous entity signals — stand out to AI assistants long before their competitors catch up. And none of the fixes require a redesign: they are surgical, testable changes to an existing website.
Citing this benchmark
You are welcome to cite these numbers with attribution: AppSynth AI Readiness Benchmark, July 2026 — appsynth.pl/benchmark. The methodology, including every check category and the scoring model, is documented on our methodology page. We re-run the benchmark when the methodology version changes, so medians stay comparable within a version.
Where does your website stand?
The free mini-audit runs 5 of the 33 checks on your domain and shows your preliminary AI Readiness Score in under a minute.
Get a free mini-audit → How we measure