Skip to content
Data report Mention check

AI mentions: 50 German online shops tested

📅 1 November 2026 ⏱ 7 min read 📊 Own measurement, raw data included

Is your shop named when a customer asks an AI? Instead of claiming, we measured: 50 German online shops across 10 sectors, 5 typical customer questions each to ChatGPT, Claude and Perplexity – 750 real API answers, collected on 1 November 2026 with the AI mention check. The brand was never part of the question. Here are the numbers, without hype.

The result in four numbers

  • 68 % of all engine answers named the shop unprompted (average over 50 shops, 15 answers each).
  • 8 of 50 shops were named in all 15 answers: bauhaus.info, bergfreunde.de, hornbach.de, mindfactory.de, notebooksbilliger.de, rossmann.de, saturn.de, zooplus.de.
  • 34 of 50 reached at least 10 of 15 mentions.
  • 9 of 50 were named three times or less, 3 of them never – including two well-known fashion retailers.

The spread is the real finding: same question, same three engines – and shops of similar size land anywhere between 0 and 15 mentions. Brand awareness alone does not explain it.

Mention rate by sector

SectorAvg. mention rateShops
Sports & outdoor 92 %5
Electronics 85 %5
Drugstore & beauty 84 %5
Household & appliances 71 %5
Pet supplies 69 %5
DIY & garden 65 %5
Furniture & living 60 %5
Food 60 %5
Toys & baby 55 %5
Fashion 35 %5

Sports/outdoor, electronics and drugstores lead – sectors with a few large specialists that have been present in tests, forums and price comparisons for years. Fashion trails: there the engines name brands and marketplaces rather than shops.

Three engines, three opinions

EngineMentionsRate
Claude (Anthropic)193 / 250 77 %
ChatGPT (OpenAI, gpt-4o-mini)153 / 250 61 %
Perplexity (sonar, Websuche)161 / 250 64 %

Claude names most generously, ChatGPT (gpt-4o-mini) most cautiously. Perplexity sits in between but answers from live search results – there, what the web says about the shop today decides. For 13 shops the engines differ by three or more mentions. Checking a single engine says little; that is why the mention check always asks all three.

All 50 shops

Shops with at least one mention are listed by name. The three shops without any mention are anonymised – a “0” next to a company name would read as a claim about the shop, while it is only a finding about three AI systems on one day.

ShopSectormentionedClaudeChatGPTPerplexity
bauhaus.infoDIY & garden15/155/55/55/5
bergfreunde.deSports & outdoor15/155/55/55/5
hornbach.deDIY & garden15/155/55/55/5
mindfactory.deElectronics15/155/55/55/5
notebooksbilliger.deElectronics15/155/55/55/5
rossmann.deDrugstore & beauty15/155/55/55/5
saturn.deHousehold & appliances15/155/55/55/5
zooplus.dePet supplies15/155/55/55/5
conrad.deElectronics14/154/55/55/5
decathlon.deSports & outdoor14/155/55/54/5
douglas.deDrugstore & beauty14/154/55/55/5
fahrrad.deSports & outdoor14/155/54/55/5
flaconi.deDrugstore & beauty14/155/54/55/5
fressnapf.dePet supplies14/155/55/54/5
mediamarkt.deHousehold & appliances14/155/55/54/5
parfumdreams.deDrugstore & beauty14/154/55/55/5
sportscheck.comSports & outdoor14/154/55/55/5
home24.deFurniture & living13/154/55/54/5
mymuesli.comFood13/155/53/55/5
toom.deDIY & garden13/155/54/54/5
zalando.deFashion13/154/55/54/5
babymarkt.deToys & baby12/155/55/52/5
bike-discount.deSports & outdoor12/155/52/55/5
bofrost.deFood12/155/52/55/5
euronics.deHousehold & appliances12/154/54/54/5
medpets.dePet supplies12/155/53/54/5
rewe.deFood12/154/54/54/5
alternate.deElectronics11/155/52/54/5
breuninger.comFashion11/153/53/55/5
baby-walz.deToys & baby10/155/52/53/5
expert.deHousehold & appliances10/154/54/52/5
jako-o.deToys & baby10/155/54/51/5
poco.deFurniture & living10/155/52/53/5
westwing.deFurniture & living10/153/53/54/5
cyberport.deElectronics9/155/50/54/5
mytoys.deToys & baby9/154/55/50/5
futterhaus.dePet supplies8/155/50/53/5
xxxlutz.deFurniture & living8/155/52/51/5
dm.deDrugstore & beauty6/153/53/50/5
koro.deFood6/152/50/54/5
wayfair.deFurniture & living4/152/52/50/5
hagebau.deDIY & garden3/152/50/51/5
obi.deDIY & garden3/152/51/50/5
zooroyal.dePet supplies3/151/50/52/5
ao.deHousehold & appliances2/152/50/50/5
gourmondo.deFood2/151/50/51/5
peek-cloppenburg.deFashion2/152/50/50/5
a fashion shopFashion0/150/50/50/5
a fashion shopFashion0/150/50/50/5
a toy shopToys & baby0/150/50/50/5

Method – and what the numbers do not say

  • Questions: 5 customer questions per shop from the fixed grid of the mention check (recommendation, value for money, shipping, comparison, getting started), phrased for the shop’s topic – e.g. “Which online shops for bicycles and bike accessories in Germany can you recommend? Name specific names.” The shop name never appears in a question.
  • Engines: Claude Sonnet 4.6, GPT-4o-mini, Perplexity sonar (with web search) – each via the official API, max. 300 answer tokens, 1 November 2026 between 10 and 11 a.m. CET.
  • Counting: mentioned = domain or brand name (including spelling variants) appears in the answer. No weighting by position.
  • Topic: set by hand for all 50 shops (e.g. “drugstore products and cosmetics”), because 16 shop home pages block automated requests – a finding that improved the mention check itself: it now detects bot-protection pages and asks for the topic instead of asking a meaningless question.
  • What the numbers do not say: whether the shop is recommended (only: named), what the number looks like next week (engines change; Perplexity daily), and why a shop is named. A sample of 50 is a picture, not a statistic. We will repeat the measurement in three months.
  • Cost of the measurement: about 2 US dollars in API fees for all 750 answers.

Check your own shop

The AI mention check asks the same questions for your domain – one question free, all five in every paid plan and in the 30-day trial. Whether the number moves after a change is what the weekly mention monitoring on aiready.tools shows. What we do not promise: that any measure guarantees a mention. Nobody can – which is why we measure.

Read on in the same cluster

🔍 How AI engines decide who gets cited →
Evidence-based: which levers really drive AI citations.