What AI assistants said about 81 Australian Shopify stores

A dated snapshot. These counts come from checks run 28 September to 1 October 2026. AI assistants answer differently from one day to the next, so read them as what we saw then, not as what they show today.

While preparing outreach for ShelfSight, we checked 81 independent Australian Shopify stores across 39 product categories. Homewares was the largest group, with 7 stores; coffee, fashion, jewellery and swimwear had 4 each. For each store we wrote 3 buying questions a shopper might ask, and asked each one once on ChatGPT, Perplexity and Gemini, set up as an Australian shopper. That is 243 questions and 729 answers. 44 answers failed (Gemini, in the first batch, before its paid tier was switched on), which leaves 685 usable answers.

We matched every answer against the store with the same rules ShelfSight uses in the app. An answer either named the store, didn't, or was unclear: it used words from the store's product titles without naming the store, so we couldn't honestly say whose product it meant. Unclear answers are counted on their own line, never treated as a miss.

The numbers

• Usable answers: 685
• Named the store: 217
• Didn't name it: 155
• Unclear: 313
• Stores named in at least 1 of their answers: 71 of 81
• Stores named in none: 10
• Stores named in 5 or more of their 9 answers: 11

By assistant:

• ChatGPT: named the store in 65 answers, didn't in 34, unclear in 144
• Perplexity: named it in 74, didn't in 76, unclear in 93
• Gemini: named it in 78, didn't in 45, unclear in 76, and 44 failed

Being named once was common here. Being named in most answers was not.

71 of the 81 stores came up at least once, which sounds reassuring. Only 11 were named in 5 or more of their 9 answers. For most of these 81 stores the picture was "sometimes": named for one question on one assistant, absent or unclear on the next. One answer on one day tells you very little, which is why ShelfSight counts answers over repeated scans instead of reporting a single check.

Which sites the assistants cited

These came up most when the assistants cited sources. Each number is how many usable answers cited the site, leaving out the 81 stores themselves:

1. etsy.com: 154
2. theiconic.com.au: 76
3. amazon.com.au: 70
4. myer.com.au: 64
5. davidjones.com: 59
6. ebay.com.au: 41
7. target.com.au: 41
8. reddit.com: 40
9. bigw.com.au: 33
10. bunnings.com.au: 33

In these answers, marketplaces and large retailers made up most of the top ten. Etsy was cited in more answers than any other site, by a wide margin, and Reddit is the only discussion site in the top ten.

Why so many answers were unclear

313 of the 685 usable answers were unclear, more than either of the other outcomes, and ChatGPT had the most: 144 of its 243 answers. In these answers the assistant used distinctive words from a store's product titles, such as a product type and a material, without naming the store or linking to its site. That could be the store's product or anyone else's. We don't guess, so these answers count neither for nor against a store.

What this means for a Shopify store

• Ask your own buying questions more than once, on more than one assistant, before drawing conclusions.
• Check your product data: clear product types, real descriptions and barcodes make it easier for assistants and shopping feeds to recognise your products.
• Look at which sites the assistants cite in your category. If marketplaces or the big department stores carry your products, check those listings are accurate; the assistants may be reading them.
• Treat a single answer as a snapshot. Counts over time are what tell you whether anything changed.

How we ran it

• 81 Australian Shopify stores, found while preparing outreach; not a random sample of all stores.
• 3 questions per store, each asked once on ChatGPT, Perplexity and Gemini with Australian location settings, through the assistants' public APIs with web search on.
• The same matching rules as the app: a link to the store's site or the store's name counts as named; title words alone are unclear. The full rules are on our methodology page.
• No store in this post is named, and the counts cover every answer we collected, not a selection.

Want the same check for your store? The free report checks your public product data with no install: run the free AI visibility report. Repeated counts on your own buying questions run in the app: Install free on Shopify.