158k AI answers studied: page tweaks explain 2.8% of citations, brand memory acts as a ranking factor
5le · x · 2026-11-16
University of Toronto researchers Mahe Chen and Nick Koudas tracked 158,099 answers from ChatGPT, Gemini and Perplexity over 90 days, then re-ran the same questions via API (arXiv:2610.15033).
Key findings:
- Page features barely matter: FAQ blocks, schema, and page speed explain only 2.8% of within-site citation differences.
- ChatGPT's own retrieval dominates: pages fetched for the model's rewritten sub-queries explain 32% of citation variance, and those sub-queries rarely match what Google or Bing return.
- Brand name works like a ranking factor: in controlled tests, swapping only the brand name (same facts) reordered product rankings — 80% top-3 rate as name A vs 39% as name B.
Takeaway: AI search citations are driven by the model's own retrieval behavior and brand memory, not traditional on-page SEO.
More from Research
- PaperILike: Human Interventions as Both Explicit and Implicit Robot Feedback — tomssilver · 2026-11-16
- EMBER gene foundation model: frozen gene representations probe well for perturbation prediction — jmuiuc · 2026-11-16
- USTC Demonstrates Autonomous Trapped-Ion QEC, Logical Qubit Lasts 11.7 ms — jiqizhixin · 2026-11-16
- Asimov Press founder: academic journal innovation is strangled by indexers like Clarivate — NikoMcCarty · 2026-11-16
- dspy-wsp Lets a Frontier LLM Edit Small Model Weights Directly: +18.3 Pts for ~$5-8 per Run — stanfordnlp · 2026-11-16
- Song Han Team Proposes Coco, a Proactive Desktop Agent with On-Device Context Observation — stanfordnlp · 2026-11-16