Which Sources ChatGPT Prefers: The Visibility Secrets Brands Need to Know

Most marketers chasing AI-search visibility track 'citations' — whether AI mentions their brand in its answer. But that's the outcome, not the cause. According to an analysis of 5 million queries collected in April 2026, ChatGPT, Perplexity and Grok break a user's question into several supporting sub-queries (a 'fan-out') before sending it to search — and it's this stage that determines, in advance, whether a brand gets cited at all.

One of the study's most actionable findings is Reddit's rapidly growing role. The share of direct Reddit references in ChatGPT queries jumped from roughly 0.15% to 3.68% between January and May 2026. That's no accident: AI systems are seeking out something polished brand content can't provide — unfiltered customer opinion and real-world usage experience.

To merge results, ChatGPT uses a method called Reciprocal Rank Fusion: content that reappears across multiple sub-queries scores far higher than content that shows up in just one. In practice, that means a brand covering a topic from several angles at once — a guide, a comparison, a case study, a Q&A — rather than through a single, even high-quality page, improves its odds of getting cited.

The practical audit process is changing too: a brand first needs to identify how high-intent queries in its category get split into sub-queries, then assess which type of source (official site, review platform, forum) gets picked for each sub-query, and finally check whether its own content shows up on those exact source types. The gap between existing content and what the AI is actually looking for is the real optimization target.

The author stresses that attempts to artificially 'manage' Reddit backfire: AI systems value authenticity and don't reward brand-orchestrated, staged discussions. The right path is building a product genuinely worth talking about, delivering strong customer service, and participating authentically in relevant communities.