DATA STUDY - What AI Cites for B2B Evaluation-Stage Prompts

Nate and Kevin dig into Ten Speed's latest research report on LLM visibility, breaking down over 7,000 citations across 220 evaluation-stage prompts to see what AI models actually cite when B2B buyers are deep in the consideration phase. The results challenge the popular narrative around Reddit and third-party content: brand-controlled content (product pages, articles, homepage) dominates citations, making up 88% of what gets cited at this stage.

From there, the conversation shifts to the practical side of LLM visibility: how to build a consistent prompt-tracking framework, why 40 to 50 prompts is the right number for most B2B brands, how to actually write prompts that reflect real buyer language instead of generic keyword-style queries, and why sentiment — not just visibility — is the metric that tells you whether your brand is being represented fairly.

Study: https://www.tenspeed.io/blog/what-ai-cites-b2b-evaluation-stage

Key moments:

0:00 – Introduction

1:01 – The Evaluation-Stage Study: Scope and Methodology

1:58 – Why Focus on the Evaluation Stage

5:15 – Study Results: Brand-Controlled Content Dominates

5:45 – The Product Page Opportunity

6:38 – Debunking the Reddit and Third-Party Content Myth

15:44 – Comparison Content Punches Above Its Weight

18:01 – The Homepage's Renewed Importance

19:32 – Building a Framework for AI Visibility Tracking

23:22 – The 40-to-50 Prompt Rule and Funnel Weighting

27:40 – How to Actually Write Good Prompts

34:41 – Brand Sentiment: When Mentions Aren't Favorable

36:04 – Taking Ownership: Reviews and Business Fundamentals

40:31 – Beyond Visibility: Connecting to Business KPIs

42:03 – Closing and Outro