When someone asks ChatGPT or Google's AI Mode for "the best waterproof dog coat for a small breed", the answer links to sources. The question for a small shop is simple: is your site one of them, or is it all Amazon and Reddit?
LLM Pulse, a company that tracks brand visibility in AI answers, published one of the larger counts so far. It looked at 391,073 citations in answers to 1,468 non-branded shopping questions ("best X", "where to buy Y", "top X for Z") asked between 12 April and 11 July 2026, across ChatGPT, Google AI Mode, Google AI Overviews, Gemini and Perplexity. Two numbers stand out. Brand, manufacturer and long-tail sites took 64.7% of the citations. Major retailers and marketplaces took 2.9%.
Who gets cited
Here is the full split by source type. Note how small the big-retailer slice is next to Reddit and YouTube. The study counts 40,678 citations for those two against 11,308 for every big-box retailer combined.
The assistants differ a lot. ChatGPT cites big retailers most often, and Gemini hardly at all:
Read it with the limits in mind
This is useful, but it is one company's sample, and the authors say so. The questions are weighted toward the categories their monitored brands sell in. It is a sample of tracked queries, not a census of US shopping questions. Two details of the method matter for a small shop:
- Brand self-citations were excluded. When the question names a brand and the answer cites that brand's site, it was not counted. The 64.7% comes from non-branded questions, which is the situation you care about: a shopper who does not know you yet.
- "Long-tail" is in the big bucket. The 64.7% mixes big manufacturers with small independent sites. The study does not say how much of it goes to stores your size, so do not read it as "small shops get two thirds of the traffic".
The honest takeaway: in these answers, the assistants mostly quote the people who make or sell the thing on their own site, plus Reddit and YouTube, and rarely the marketplace listing. Your own product pages are in the running. A marketplace listing alone will rarely get you named.
What your product pages need
The study counts citations. It does not test which page features earn them. So treat the list below as the basics that make a page usable as a source, not as a guaranteed ranking recipe.
- Answer the question the shopper asked. "Best coat for a small breed" needs sizes by breed or chest measurement, not just S/M/L. Put the deciding facts in plain text near the top: what it is for, who it suits, who it does not, sizes, materials, care.
- Write comparison facts, not adjectives. "Seams taped, tested in two hours of rain" can be quoted. "Premium quality" cannot. If you ran a test yourself, say how.
- Keep price and stock visible and correct. An answer that says "in stock, $49" and lands on a sold-out page costs you the sale and the trust.
- Add product structured data. Most shop platforms already output it. Check that yours includes price, currency and availability. Paste a product URL into Google's Rich Results Test to see what it reads.
- Make sure the text is in the HTML. If sizes and descriptions only appear after a script runs, some crawlers will not see them. View the page source and search for your size table.
- Write one honest buying guide per category. A page such as "How to choose a dog raincoat" answers the "best X for Y" question directly, and can link to your products where they fit.
- Do not block the search crawlers you want quoting you. Check robots.txt for the bots behind the assistants in the study.
A robots.txt that allows the search crawlers for ChatGPT, Perplexity and Google looks like this (if your file already has a blanket Disallow: / for unknown bots, these explicit groups are what lets them in):
User-agent: OAI-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Googlebot
Allow: /
Sitemap: https://example-shop.com/sitemap.xml
And a minimal product block, if your platform does not output one, goes in the page's <head>:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Product",
"name": "Packable dog raincoat, small breeds",
"description": "Taped seams, reflective trim, sizes by chest measurement from 30 to 50 cm.",
"sku": "RC-SM-01",
"brand": {"@type": "Brand", "name": "Example Shop"},
"offers": {
"@type": "Offer",
"price": "49.00",
"priceCurrency": "USD",
"availability": "https://schema.org/InStock",
"url": "https://example-shop.com/products/dog-raincoat-small"
}
}
</script>
Reddit and YouTube: worth showing up, not worth faking
Reddit and YouTube together took 10.4% of citations, more than every big retailer combined. If customers already talk about your products there, that helps. Faking it does not: planted posts get removed and remembered. The cheap, honest version is a short video of the product in use on your own channel, and answering questions in threads where you are genuinely the maker (say so).
A 30-minute check for this week
- Ask ChatGPT, Perplexity and Google AI Mode three non-branded questions your customers would ask. Write down who gets cited.
- Open your best-selling product page and check: deciding facts in plain text, price and stock correct, structured data present.
- Open
/robots.txtand confirm you are not blocking the search crawlers above. - Repeat the three questions in a month. Without a tracking tool, that is the only measurement you have, so keep the notes.
