How was the audit run?
The audit scored the 25 highest-traffic Danish online retailers on one 10-point rubric, locked before any site was opened. The sample was ranked by SimilarWeb's Denmark e-commerce rank and cross-checked against Dansk Erhverv and FDIH. Each check is binary, worth 10 points, and 70 or higher is "ready."
Two choices matter. Product and review markup were checked on real product pages, not homepages, because 0 of the 21 homepages we fetched carried Product schema. And every check reads raw server HTML only, never JavaScript, because the major AI crawlers (GPTBot, ClaudeBot, PerplexityBot) do not run JavaScript either, so client-side schema is invisible to them and scores as absent.
| # | Check | Verification |
|---|---|---|
| 1 | AI crawlers not blocked in robots.txt | Verified (raw) |
| 2 | Organization or Store schema present | Verified (raw) |
| 3 | Product schema with price and availability | Verified (raw) |
| 4 | Entity clarity (About page, brand, contact) | Editorial |
| 5 | Citable facts as text, not images | Editorial |
| 6 | Review or rating markup present | Verified (raw) |
| 7 | Answer-shaped FAQ content | Editorial |
| 8 | Low JavaScript dependency or llms.txt | Proxy |
| 9 | Surfaces for a category buyer query | Retrieval proxy |
| 10 | Trust signals (HTTPS, contact, policies) | Editorial |
Which stores publish review markup AI search can read?
Only 5 of 25 do: Matas, JYSK, POWER, Imerco and Apopro. The other 20 display stars to shoppers but never tag them for machines. Reviews are among the strongest trust signals a shop can offer for citation, and the GEO research of Aggarwal and colleagues (KDD 2024) found statistics and cited sources give the biggest visibility lift. Marking up reviews you already collect is a few hours of work. One caveat: Apopro's rating is store-wide, not product-level, which is worth disclosing.
How many publish Organization and Product schema?
Better than reviews: 14 of 25 publish the Organization schema that tells an engine who they are, and 21 of 25 publish Product schema with price and availability once you check a product page, not the homepage. The four without it split cleanly. Rema 1000 and Coop have no public product catalogue (grocery), while Plantorama and ILVA render product data in JavaScript, invisible to AI crawlers.
Does any retailer opt out of AI search entirely?
Yes, one. thansen.dk blocks GPTBot, ClaudeBot and Google-Extended in its robots file, telling the major AI engines not to read the site at all. That is not a gap a few hours of markup fixes. It is a deliberate choice to be invisible to AI search.
Are the biggest stores the most ready?
Not by size, but by publishing choices. 23 of 25 clear the 70 readiness line, the field averages 81 of 100, and only Coop (50) and the borderline Proshop (60) fall below it. JYSK and Apopro hit a perfect 100. Proshop's low score is partly an artifact: its Cloudflare wall blocks crawlers, yet its archived HTML does carry Product schema. The same caveat covers Zalando, Elgiganten and the access-gated Nemlig, all verified from archived or Googlebot-served HTML and reported as indicative, not exact.
| # | Store | Category | Org | Product | Review | Score | Ready |
|---|---|---|---|---|---|---|---|
| 1 | Zalando | Fashion | no | yes | no | 70* | yes |
| 2 | jem & fix | General | yes | yes | no | 90 | yes |
| 3 | Matas | Beauty & health | no | yes | yes | 90 | yes |
| 4 | Harald Nyborg | General | yes | yes | no | 90 | yes |
| 5 | Rema 1000 | Grocery | yes | n/a | no | 80 | yes |
| 6 | Elgiganten | Electronics | no | yes | no | 70* | yes |
| 7 | Proshop | Electronics | no | yes | no | 60* | no |
| 8 | JYSK | Home & interior | yes | yes | yes | 100 | yes |
| 9 | Bilka | General | no | yes | no | 80 | yes |
| 10 | Silvan | Home & interior | no | yes | no | 80 | yes |
| 11 | Coop | Grocery | no | n/a | no | 50 | no |
| 12 | thansen | Auto & leisure | yes | yes | no | 80† | yes |
| 13 | POWER | Electronics | no | yes | yes | 70 | yes |
| 14 | Imerco | Home & interior | no | yes | yes | 90 | yes |
| 15 | Nemlig.com | Grocery | yes | yes | no | 80* | yes |
| 16 | Plantorama | Home & interior | yes | js | no | 80 | yes |
| 17 | Bog & idé | Books & media | yes | yes | no | 90 | yes |
| 18 | ILVA | Home & interior | yes | js | no | 80 | yes |
| 19 | Saxo | Books & media | no | yes | no | 80 | yes |
| 20 | Salling | General | no | yes | no | 80 | yes |
| 21 | XL-BYG | Home & interior | yes | yes | no | 80 | yes |
| 22 | Med24 | Beauty & health | yes | yes | no | 90 | yes |
| 23 | Apopro | Beauty & health | yes | yes | yes | 100 | yes |
| 24 | BabySam | Toys | yes | yes | no | 80 | yes |
| 25 | Coolshop | General | yes | yes | no | 90 | yes |
* Behind a bot-protection or access wall; structured data verified from archived or search-engine-served HTML, so the score is indicative. † thansen.dk scores well but blocks the major AI crawlers in robots.txt, so its readiness is moot. "js" marks product data rendered client-side and therefore invisible to non-rendering AI crawlers.
Why is this an opportunity, not a crisis?
Because the engines already favour ecommerce, and the stores AI names in every query (Matas, Proshop, Nemlig) are named for brand authority, not markup. This echoes our Danish GEO Benchmark 2026: readiness alone barely predicts AI citations (r = 0.12). Markup is table stakes; brand authority is what gets you chosen. Readiness will not get you recommended on its own, but it removes the reasons an engine has to skip you, and it is the half of the equation you fully control.
What should a retailer do?
Treat machine-readability as table stakes, then build the authority that gets you chosen. Four controllable moves:
- Mark up the reviews you already collect, the cheapest upgrade in this dataset.
- Keep Organization and Product schema in the raw HTML, not injected by JavaScript.
- Write delivery, returns and sizing facts as real text, not images.
- Keep the AI crawlers allowed, unless invisibility is a deliberate choice.
None of it guarantees a recommendation. All of it improves your odds of getting one. If you want to see where your own store stands, Boas Media runs these exact checks and fixes what an AI engine cannot read.
"Our benchmark showed that being technically ready does not, on its own, get a Danish business cited by AI search, because brand authority is what makes an engine choose you. But readiness is the table stakes underneath that, the baseline that gives you a chance of being cited at all, and in retail the cheapest piece of it is the one almost nobody uses: only 1 in 5 of the biggest stores mark up the customer reviews they already collect. That is a few hours of work standing between a store and being quotable." Silas Bondesen, Founder, Boas Media
Note on rigor (for editors)
Four findings rest on objective, binary checks verified against each retailer's live or archived raw page source: review or rating markup (5 of 25), Organization schema (14 of 25), Product schema with price and availability (21 of 25, on product pages), and the thansen.dk AI-crawler block. Three retailers (Zalando, Proshop, Elgiganten) sit behind bot walls and one (Nemlig.com) is access-gated; their data was verified from archived or search-engine-served copies and their composite scores are indicative. The six softer rubric checks carry lower confidence, so the public claims lead on the four verified findings. This study does not claim readiness predicts AI recommendation; the companion Danish GEO Benchmark 2026 found it does not (r = 0.12). The two studies use different samples, so their per-metric figures are independent. Full method and per-store scores are free to cite under CC BY 4.0.