Blog
Research

I analyzed 250M AI responses. Here is the uncomfortable truth about AI Search right now

J

Josh Blyskal

5 Dec, 2025

At Tech SEO Connect in Raleigh last week, I shared an analysis of 250 million AI responses collected over the past 18 months. Here's the presentation translated into a post, along with some additional context that didn't fit the deck.

Uncomfortable truth #1: AI is confident even when it's wrong

AI models consistently present brand information with high confidence regardless of accuracy. We found factual errors in roughly 23% of brand descriptions across major AI platforms. The error rate is higher for newer or smaller brands and for fast-moving categories like fintech and healthtech.

Uncomfortable truth #2: Your competitors can shape how AI describes you

AI models synthesize information from across the web, including content published by your competitors. We found instances of competitor content influencing AI-generated brand descriptions. Brands that don't actively monitor and manage their AI presence are ceding narrative control to anyone who publishes about them.

Uncomfortable truth #3: SEO rankings and AI citations are only weakly correlated

The correlation between Google Search ranking and AI citation frequency is weaker than most marketers assume — around 0.4 in our dataset. Ranking #1 for a keyword does not guarantee citation in AI answers for related prompts. AEO requires dedicated effort beyond SEO.

What to do about it

  • Audit your AI-generated brand descriptions on each major platform monthly
  • Build a dedicated AEO content strategy that targets AI citation, not just search ranking
  • Use structured data and clear factual statements to give AI reliable information to draw from
  • Monitor competitor content that mentions your brand — you need to know what's shaping your AI narrative

Start optimizing your AI visibility

Join thousands of brands tracking their presence across all major AI platforms.

Get started free