Nearly one in five Swedish online stores was not mentioned at all by ChatGPT or Claude

If you ask an AI service where to buy something, does your local online store show up? Not always. That is the finding of a study from the Swedish digital marketing agency KLOTA AB in Borås.

On November 4, 2026, the agency tested 189 Swedish online stores with annual revenue of SEK 3-30 million. The starting pool was 205 companies, but 16 were excluded because they were not online stores or did not have a Swedish version.

How the test was done

For each store, KLOTA asked three purchase questions in Swedish. They were worded based on the store's homepage and did not mention the store's name. The questions went to both ChatGPT and Claude with web search. In total, 1,123 responses were analyzed.

Many were overlooked

28% of the stores were never mentioned in ChatGPT's three responses. 29% were never mentioned by Claude, and 19% were not mentioned by either. Claude mentioned the stores in 47% of responses, ChatGPT in 39%. Only about 60% of the same stores appeared in both services.

When a store was not mentioned, the AI services recommended one of Sweden's 100 largest online stores instead in four out of ten cases.

Text and questions appear to help

Stores with fewer than 220 words on the homepage were mentioned in an average of 2.4 of 6 responses. Stores with more than 970 words were mentioned in 3.3 of 6. Stores with at least one heading phrased as a question were mentioned in 3.4 of 6 responses, compared with 2.6 for stores without.

Among the 50 stores with at least 500 words and question headings, only 3 (about 6%) were never mentioned. Stores with minimal text and no question headings were never mentioned in 26% of cases.

Anders Karlsson of KLOTA says: "The door is open. What separates the stores that get mentioned from those that don't is whether there is text on the page that answers the customer's question."

Access was rarely the problem

98% of the stores allowed AI crawlers access. These are programs that automatically scan websites. However, 40% lacked specific information about shipping, delivery or returns.

About a third of the companies have an llms.txt file, a special file intended for AI services. They were not mentioned more often than those without one, and KLOTA advises against prioritizing it.

Recommendations and caveats

KLOTA recommends content that answers customers' questions, clear terms of purchase, company and contact information, checking robots.txt, and repeated tests across multiple AI services.

The study was conducted by an agency that sells AI visibility services, and it has not been independently reviewed by other researchers. The results show correlations, not cause and effect. AI responses vary from one occasion to the next, and the measurement is from a single point in time. The sample covers only Swedish online stores with Swedish content and prices in SEK, so it may not reflect Swedish e-commerce as a whole.