ChatGPT answers favor specific brands before searching

The research by Suganthan Mohanadasan, published August 10–27, 2026 in articles including "ChatGPT Already Knows Who's In The Running Before It Searches", identifies a structural bias in how generative AI systems deliver recommendations. Mohanadasan tested this by asking ChatGPT open-ended questions about product categories—such as "What is the best AI note-taking app for 2026?"—and then comparing which brands appeared in ChatGPT's initial search query, which web pages ChatGPT retrieved, and which brands ended up in the final answer the user saw.

The pattern was striking: across 27 conversations spanning software and AI tools categories, the same 7 note-taking tools appeared early in nearly every search, and those early mentions correlated strongly with final-answer inclusion. When the body restricts the analysis to software and AI-tool categories alone (21 of the 27 conversations), the relationship became even tighter: 11 out of 13 categories showed the bias. Crucially, Mohanadasan notes that these early-appearing brands were not explicitly chosen by the user; they represent a pre-selected set that ChatGPT appears to have "in mind" before the web search happens.

This mirrors the historical dominance of SEO (Search Engine Optimization), where websites learned to rank higher on Google by optimizing for the search engine's algorithm. The analogy suggests that as generative AI systems replace traditional search as a discovery channel, the same kind of ranking leverage and marketing manipulation could emerge—now called GEO (Generative Engine Optimization). The body emphasizes that the challenge for web publishers is steep: if a page does not appear in ChatGPT's early results, the odds of being cited in the final answer drop dramatically.