Short answer first: AI recommends a brand when other people vouch for it, and it cites a brand when the brand's own site gives it clear facts to quote. Most Chinese brands going global have done the first part. Very few have done the second.
That was my main takeaway from October 23, when our team spent the day at the 2nd GEO Conference in Shenzhen, hosted by Cross Border Digital.
Our CEO Eva was on the Reddit roundtable in the afternoon. The room was full of search platforms, SEO agencies, GEO providers and localization teams, and nearly every session ended up at the same question: when overseas buyers ask AI before they buy, how does your brand end up in the answer?
A few numbers from the day that I keep coming back to:
- One cross-border agency said a leading brand they interviewed already gets 7% of its revenue from AI traffic. Small, but it's the traffic closest to a purchase.
- In Western markets, Chinese brands show up in 47–57% of AI search results, yet only 7% of the traffic from AI and search actually lands on their own websites.
- Semrush research: teams that run SEO and AI search as one strategy saw positive results 86% of the time. Teams that run them separately: 31%.
That second number is the one that bothers me. We're being mentioned. We're just not the page being cited. Here's how I'd explain it to a founder over coffee. AI looks at two things.
One is trust: press, reviews, what real users say in communities. That decides whether it's willing to recommend you. The other is facts: clear product, comparison and FAQ pages on your own site. That decides whose page it quotes when it does. Chinese brands have spent years on the trust side. The fact side is usually thin, so AI recommends you and then cites a marketplace or a review site instead.
The second thing I heard again and again: this rarely fails because of method. It fails because nobody owns it. Question data sits with marketing, the website with engineering, media with PR, community with ops. Teams check "did AI mention us this week?" from their own logged-in accounts, which skews the results anyway. Before any content work, the most useful thing we usually suggest is boring: give the outcome one owner, and set a baseline of buyer questions you can re-test.
On Reddit, Eva's point on the roundtable was simple. Reddit, and communities like X and LinkedIn, matter to AI because the discussions are real and specific. Your website says who you are. User discussions say whether it actually works, and AI uses the second to check the first. A good thread also keeps getting found and cited long after it's posted, unlike an ad that disappears when spend stops.
One caution we always give: brands that register an account and start posting usually get removed fast. It has to be treated as long-term work, with respect for each community's rules. One line from the localization panel stuck with me: "Translation gets you understood. Resonance gets you chosen." I'd add that now local AI has to understand you too, and a machine-translated website doesn't get you there. Real local discussion does. So, three direct answers: Why does my brand appear in AI answers but get little traffic? Because AI is citing third-party pages, not yours. You have trust but no citable facts on your own site. How do I get cited by ChatGPT? Give it something to cite on your site, then make sure media, reviews and community discussions say the same thing and point back to it.
Does Reddit content get used by AI search? Speakers at the conference said mainstream models tend to cite it, and threads that answer recurring questions keep being cited long after posting. For what it's worth, the clearest proof I've seen of this is one DTC brand we work with: 23 pieces of content that AI engines now cite often, and organic channels driving about 30% of its sales. The race isn't about who shows up most in AI answers anymore. It's about who becomes the answer. If you're working on this for a brand going global, I'd genuinely like to compare notes.