- AI cites third-party sources over brand websites — mainly video, review sites, and forums. The top sources across 165 answers were YouTube (208 citations), RunRepeat (129), and Reddit (52).
- Which source wins depends on the engine. ChatGPT leaned on editorial sites like Tom’s Guide and WhoWhatWear and almost never cited YouTube or Reddit; Gemini and Perplexity did the opposite, citing YouTube more than any other single source.
- Being named by AI and being cited as a source are different things: Nike was named in 93% of answers but nike.com was cited just twice.
- Perplexity cited a brand’s own website zero times, and X/Twitter got zero citations across all three engines — AI leans on video and community sources instead of brand or social content.
How was this sneaker experiment run?
We tracked 5 buyer-intent sneaker prompts — the kind someone types right before buying, not generic research questions — through ChatGPT, Gemini, and Perplexity once a day for 11 straight days (June 26 to July 6, 2026). Every run, we logged every brand the AI named and every source it cited.
That produced 165 answers, 1,656 citations, and 588 distinct brands named.
The exact 5 prompts we tracked:
- What are the best sneakers for everyday streetwear outfits in 2026 that are comfortable for walking all day and also look stylish?
- I want sneakers that are both good for light running and also match smart casual outfits. What types of shoes or brands should I look for in 2026?
- I’m looking for sneakers that are comfortable for walking 10,000+ steps a day and still look stylish with jeans and oversized outfits. What are the best brands in 2026?
- What are the best sneaker and athletic shoe brands in 2026 for both performance and fashion? How do they differ in design style and use cases?
- Compare Nike, Adidas, and On sneakers in terms of design, fashion appeal, comfort, and performance. Which brand is more lifestyle-focused vs performance-focused?
We watched four brands closely: Nike (the own brand), plus Adidas, On, and Puma as competitors. For every answer we recorded two separate things: whether the brand was mentioned in the text, and whether a page on the brand’s own domain was cited as a source.
Those two numbers turned out to tell very different stories (leadrescue.app analysis, 2026).
What is the difference between being mentioned and being cited?
A mention is when an AI names your brand in its answer. A citation is when it links to a page as the source behind that answer.
They are not the same, and the gap can be enormous. Nike was mentioned in 153 of 165 answers (93%) but its own website was cited in just 2 of them.
Adidas and On were mentioned slightly less often, yet their own sites were cited far more: 11 and 12 times. Puma barely registered at all, named in about 22% of answers with zero citations.
So the pattern is not simply “big brands get cited.” Nike, the most-recommended brand of the four, had the least-cited website by a wide margin.
Where does AI actually get its information about brands?
From third-party publishers and communities, not brand-owned pages. Across our 165 answers, the most-cited sources were YouTube (208 citations), RunRepeat (129), and Reddit (52).
Review sites and news outlets filled out the rest. No brand’s own website ranked near the top of the list.
The contrast is stark once you rank every source. Of 245 distinct domains the engines cited, nike.com sat around 100th with just 2 citations, even though Nike was the single most-named brand.
This lines up with published GEO research: a Princeton-led study found that structure, statistics, and citations lifted a source’s visibility in AI answers by up to 40% (Aggarwal et al., Princeton, 2023). Independent, evidence-shaped pages get pulled; brand marketing pages mostly do not.
Why does AI recommend a brand but not cite its website?
AI recommends brands without citing their websites because it prefers independent, third-party validation over self-promotional marketing. It usually cites authoritative review sites (like RunRepeat), forums (like Reddit), or media articles to prove a brand is actually trusted by real users.
A brand’s own site reads as marketing; third-party coverage reads as evidence. That is why a household name like Nike gets recommended from memory and third-party coverage, while its own pages stay uncited.
A generative engine is an AI system like ChatGPT or Perplexity that writes an answer instead of returning a list of links. To back a claim, it reaches for sources it can quote and trust, which skews toward reviewers and communities.
In our data, the effect was strongest on Perplexity, which cited a brand’s own website exactly zero times across all 165 answers. Sentiment for Nike was still overwhelmingly positive (124 positive mentions to 1 negative), so this is not about the brand being disliked.
It simply is not the source.
Which platforms do AI engines cite most?
Video and forums dominate; social posts barely register. YouTube led every source with 208 citations, followed by review sites and Reddit (52).
TikTok, Facebook, and Instagram appeared only through Perplexity. And X/Twitter was cited zero times across all 165 answers and all three engines, a striking gap for a platform that size.
| Platform | Citations | Which engines cited it |
|---|---|---|
| YouTube | 208 | Gemini, Perplexity |
| 52 | All three | |
| TikTok | 23 | Perplexity only |
| 19 | Perplexity only | |
| 14 | Perplexity only | |
| Quora | 5 | Gemini |
| X / Twitter | 0 | — |
The likely reason X is absent: its content is harder to crawl and is not where durable product-review content lives. Video reviews and Reddit threads are, and the engines lean on them heavily.
Which platform does each AI engine source from the most?
The table above blends all three engines together, but each one actually has its own sourcing habits. ChatGPT cited YouTube zero times and Reddit just once across its 254 citations, leaning on editorial sites like Tom's Guide instead.
Gemini and Perplexity did the opposite, citing YouTube more than any other single source. Gemini cast the widest net of the three (928 citations across 185 distinct domains), while Perplexity repeated a narrower set of sources most often (474 citations across just 49 domains).
| Engine | Top source | 2nd most cited | 3rd most cited |
|---|---|---|---|
| ChatGPT | Tom's Guide (44) | WhoWhatWear (31) | REI (11) |
| Gemini | YouTube (107) | RunRepeat (96) | WhoWhatWear (53) |
| Perplexity | YouTube (101) | Reddit (42) | NYTimes (31) |
The practical takeaway: earning a mention on YouTube or Reddit mainly helps you with Gemini and Perplexity. If ChatGPT is the engine sending you the most traffic, put that same effort into editorial and review sites like Tom's Guide and WhoWhatWear instead, since that is where ChatGPT is actually pulling its answers from.
What does this mean for getting your brand cited by AI?
Earn coverage on the sources AI already trusts. If engines cite YouTube, review sites, and Reddit, then reviews, video, and community mentions do more for your AI visibility than polishing your homepage.
Being named is not enough. You want to be the brand whose evidence trail the engines can actually find and quote.
- Get independently reviewed. A single strong review on a site like RunRepeat or Tom’s Guide is worth more to AI than a page of your own copy.
- Show up in communities. Reddit was cited 52 times. Genuine, helpful presence in the right threads feeds the engines.
- Treat each engine separately. Perplexity never cited a brand’s own site, so leaning on your domain alone leaves you invisible there. This is the same per-engine split we found when we asked whether ChatGPT, Perplexity, and Gemini recommend the same brands.
This is also why measuring mentions alone hides the real picture. You need to see which sources each engine cited for your prompts, which is exactly what Lead Rescue’s citation report tracks across ChatGPT, Gemini, and Perplexity.
To learn the metric behind it, start with what AI brand visibility means, or see the full method in how to track AI search citations.
Lead Rescue shows you every source each engine cited for your brand, so you can go earn the coverage that gets you into the answer. See where AI engines get your brand →