AI learns about your brand from other people first
You spent years on your website, the copy, the story, the product pages. AI barely read it. When a shopper asks about your brand, AI answers with what other people wrote about you, on sites you don't own and can't edit. It's repeating them, not you.
We ran a full report on Quince, the direct-to-consumer apparel brand, across ChatGPT, Gemini, and Claude, and traced every source behind what the models said. Here's what the source list showed.
- 93% of the citations behind AI's answers about Quince came from sites Quince doesn't own.
- The sites doing the talking are sustainable-fashion guides, independent style blogs, and Reddit threads, not quince.com.
- Those same sites are already naming Everlane, Reformation, and Madewell. The pages that decide your visibility are deciding your competitors' too.
Where does AI actually get its answer about you?
Where AI gets its answer about you is mostly off your own site. Across the 5,661 citations behind AI's answers to 219 buyer questions about Quince, 93% pointed somewhere other than quince.com.
This isn't a knock on Quince. AI clearly knows the brand and likes it: it reinforces the core story without being asked, calling Quince "affordable luxury" 298 times and crediting its factory-direct model 206 times. The positioning lands. The point is where that impression comes from. When a shopper asks an assistant whether Quince is worth buying, the answer is built mostly from pages Quince never wrote and can't edit.
That ratio is the part most owners get backwards. You spend your effort on the homepage, the product copy, the brand story. AI spends most of its attention somewhere else.
Which sites is AI actually reading?
The sites AI is reading are mostly editorial guides, independent blogs, and community threads. For Quince, the most-cited outside sources were ecocult.com and The Good Trade (sustainable-fashion guides), apartstyle.com, danaberez.com, and closo.co (independent style writers), Reddit threads, and a little mainstream press from Business Insider.
Notice what's missing from that list: ad campaigns, the brand's social feed, its press releases. AI didn't reach for any of it. It reached for the places where someone other than the brand had already written something specific and useful, a comparison, a review, a "best affordable cashmere" roundup. Those are the pages it trusts to answer a buying question, because that's what they were written to do.
So the real question isn't "what does my website say." It's "which of these pages have ever mentioned me, and what do they say." That list is short and concrete, and you can go work on it.
Why is AI quoting them instead of your own site?
AI quotes other sites instead of yours when your own site is hard for it to read. The Quince report flagged three fixable gaps: no structured product data, no FAQ content, and no llms.txt file telling AI which pages matter. With those missing, the models can't pull clean facts straight from the source, so they lean on whatever third-party pages they can parse instead.
The report put it plainly: add structured data to your product pages and "AI quotes your tagged product pages instead of pulling from review sites." Right now the review sites win by default. A machine-readable site doesn't replace third-party coverage, but it does get your own words back into the answer, where you have some control over them. We found the same pattern across our broader study of how AI picks brands: the sites AI can read easily are the ones it cites.
These same sites are choosing your competitors too
The same sites that decide whether AI names you are already naming your competitors. On the pages AI cited most for Quince, the brand almost never appears alone. ecocult, The Good Trade, apartstyle, closo, and thecommons all name Everlane, Reformation, and Madewell in the same breath, often ahead of Quince.
A handful of sustainable-fashion guides and style blogs are quietly shaping the shortlist for an entire category. Earn a spot in those pages and you show up next to the brands AI already favors. Stay off them and you're invisible no matter how good your own site looks. The competitive shelf isn't just on Google anymore. It's spread across two dozen sites you don't own, and the brands that earn coverage there get recommended.
This is what a source map gives you that a traffic report never could: not "how many people visited," but "which specific pages are teaching AI what to say about my category, and am I on them."
See the sites AI reads about you
Want your own list? Run the report on your business and it traces every source behind AI's answers about you, grouped by type, plus the sites citing your competitors that never mention you. The $39 full report is the one that maps your sources. It's in your inbox in under an hour, with no account and no sales call. The 93% you don't control is the part you can actually go win.