No Norwegian online store has FAQ schema. Not one.

What did the study find?

The audit in Online Visibility Report measures a separate category for citability: whether the content is shaped so an answer engine can lift it out and reuse it. Among the 78 sites the crawler could read:

• 0 of 78 have FAQPage, QAPage or HowTo schema, machine-readable question-and-answer markup
• Only 15% pass the test for question-shaped headings (72% get a warning)
• Only 22% pass the test for concrete evidence, numbers, sources and fact-based paragraphs in the text
• 56% lack structured data (JSON-LD) entirely. Among those with something, we found WebSite on 37%, Organization on 31% and BreadcrumbList on 5%, and in the machine-extracted sample we found no Product or Offer schema anywhere

The contrast with the basic technical categories is striking: crawlability averages 95 out of 100, mobile 90. Norwegian online stores are easy for machines to read, and nearly impossible for them to cite.

Why does question form matter so much?

Because it's this that seems to separate the winners from the losers in AI answers, more than size, age and authority.

In the study we looked closer at three segments and compared direct competitors. In all three pairs, the store with the lowest domain authority won the battle for AI citations:

• Sports: Milrab (authority 2.8) is cited 3 of 3 times the AI Overview appears. XXL (4.7): 0 of 3.
• Fashion: Miinto (3.1) is cited 4 of 5. Boozt (5.7): 0 of 5.
• Pharmacy: Farmasiet (4.7) is cited 3 of 3. Apotek 1 (5.6): 2 of 5.

What sets the winners apart isn't a better technical score: Boozt and Milrab have an identical AEO subscore in the audit (44 of 100), despite a 0% versus 100% citation rate. The difference lies in concrete structural choices: content that's already shaped as a finished answer. Question-phrased headings. Short, fact-based paragraphs. FAQ structure. Boozt's report notes that the site “lacks the necessary content structure and quotability,” while Milrab's report describes it as a “reference player for AI-generated answers” in its niche.

An honest caveat here: the head-to-heads were studied in three segments, not all 99. But the pattern is consistent where it was checked, and it points the same way as the overall picture: domain authority correlates with virtually nothing in AI citation rate (r = 0.05).

And that's exactly what makes 0 of 78 good news, if you run an online store: authority is built over years, but content form can change in days. The field is wide open. First come, first served.

What actually is FAQ schema?

Two things, which belong together:

1. The visible content: a section on the page where real customer questions get short, concrete answers.
2. The markup: a small JSON-LD block in the page's source code that tells machines “this is a question, this is the answer,” so the content can be extracted without guessing.

One important clarification, so we don't promise something that isn't true: FAQ markup no longer earns star treatment in regular Google results. Google effectively removed FAQ snippets from search results for most sites in 2023. Today the value lies elsewhere: with the answer engines and AI crawlers that need to understand and reuse your content, and with the customers who actually get answers to what they're wondering about. The visible FAQ content is the raw material itself; the markup makes it machine-readable.

How do you do it in practice?

1. Find the actual questions. Don't invent them in a meeting room. The customer service inbox, the site's search log, the “People also ask” boxes in Google and forum threads in your niche are full of the wording customers actually use. That's the wording AI answers respond to.
2. Write answers that can be lifted out. Two to four sentences per answer, with concrete numbers where you have them: price, delivery time, return window, stock status. That's what the evidence test measures, and only 22% pass it today. Marketing language (“fast delivery!”) isn't an answer; “delivery in 1–3 business days, free shipping over 500 kroner” is.
3. Turn the headings into questions. “Shipping and delivery” becomes “How fast do you deliver?” “Returns” becomes “How do I return an item?” This costs nothing, can be done today, and only 15% of Norwegian online stores have done it.
4. Add the markup: a JSON-LD block in <head> or at the bottom of the page, mirroring the visible content.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How fast do you deliver?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "We ship all orders within 24 hours on weekdays. Delivery takes 1–3 business days across Norway, and shipping is free for orders over 500 kroner."
      }
    },
    {
      "@type": "Question",
      "name": "How do I return an item?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "You have a 30-day right of return. Register the return in My Page, use the enclosed return label, and you'll be refunded within 5 business days after we receive the item."
      }
    }
  ]
}

The non-negotiable rule: the markup must mirror content that's actually visible on the page. Schema with no visible counterpart is the road to trouble, not to citation.

Cover the product and category pages too. That's where buyer questions actually land: “does this bike work for winter use?” belongs on the product page, not just on a customer service hub. And since Product and Offer schema is effectively absent from Norwegian e-commerce, that's another opportunity in the same stroke.

Feel free to try it yourself, before and after: google two or three of the questions your own customers ask, see who the AI answer cites today, and check your own site with a schema validator to see what the machines actually find there.

An honest caveat

FAQ schema is no guarantee of citation. Overall technical score correlates only weakly with actual AI citation in our dataset (r = 0.20 among the 78 we could read), and as the Boozt/Milrab pair shows, not even the AEO subscore captures the difference: it's the sum of the concrete choices that seems to work, not one single markup tag in the source code. And unlike the five configuration fixes we've written about before, this is content work: it requires someone to actually write good answers. That's why FAQ schema was deliberately kept out of the quick-win simulation in the study: it belongs in its own, larger investment category.

But that's also the whole point. Configuration is something anyone can copy in an afternoon. Good, question-shaped content with real numbers is harder to copy, and right now, none of the 78 measured stores have done it.

Closing thoughts

Three numbers to take away: 0 of 78 have FAQ schema, 15% have question headings, 22% have citable evidence. In a market where more than half of buyer questions are already answered by an AI at the top of the search results, this is the most wide-open opportunity we found in the entire dataset.

The bar is at ground level. Start with one page, ideally the customer service page you already have, rewrite the headings as questions, tighten the answers to concrete numbers, and add the markup using the example above. Then take on the product pages over time. Good luck!