AI Assistant Bias Revealed: Here’s Who Gets Favored & Who Gets Shunned in the New Era

Posted on June 19, 2025
Source: Data & Studies – SEO Blog by Ahrefs by Patrick Stox. Read the original article

TL;DR Summary of How AI Assistants Are Shaping Web Visibility Differently Than Traditional Search

AI assistants select website mentions based on criteria distinct from traditional search rankings, leading to noticeable biases. Platforms like Google AI Overviews favor user-generated content such as Reddit and YouTube, while under-representing authoritative sources like Wikipedia and major health sites. Among AI systems, Perplexity maintains the most balanced citation patterns, whereas Google and ChatGPT show stronger biases. Understanding Mention Share versus Impression Share reveals how these models influence web visibility in new ways.

Optimixed’s Overview: Understanding AI Assistant Biases and Their Impact on Online Content Visibility

Introduction to AI Assistant Influence on Web Mentions

AI assistants are fundamentally altering how websites gain visibility by citing sources differently than traditional search engines. Instead of purely counting mentions, this new approach compares Mention Share—the frequency a domain is cited—with Impression Share—the expected visibility based on topic popularity and search volume. This methodology uncovers systemic biases in AI-generated content references.

Key Findings on Source Representation

  • Google AI Overviews disproportionately cite user-generated content (UGC) sites such as Reddit, Quora, and YouTube, while underutilizing trusted informational sites like Wikipedia, Mayo Clinic, and Cleveland Clinic.
  • ChatGPT under-represents Wikipedia and major news outlets like Reuters, indicating a bias against some authoritative sources.
  • Perplexity exhibits the most neutral stance, closely aligning mention patterns with actual topic popularity and traditional web visibility.

Definitions for Clarity

  • Mention Share: The percentage of AI responses that mention a particular domain.
  • Impression Share: The weighted potential visibility of a domain based on search volume for relevant queries.
  • Potential Reach Share: An estimate of exposure a domain could have within AI assistant responses, based on search volume proxies.

Examples of Over- and Under-Representation

  • Over-relied upon by Google AI Overviews: Reddit (+3.4%), Quora (+2.2%), YouTube (+2.0%) compared to expected visibility.
  • Under-utilized by Google AI Overviews: Wikipedia (-3.2%), Mayo Clinic (-1.6%), Cleveland Clinic (-1.6%).
  • ChatGPT Over-reliance: Google.com and Apple.com slightly over-represented; Wikipedia and Reuters notably under-represented.
  • Perplexity’s Balanced Approach: Slight over-reliance on YouTube and Apple.com but generally aligns well with expected visibility, under-utilizing Wikipedia only modestly.

Implications for SEO and Content Strategy

The fragmentation of search visibility across AI assistants demands a reconsideration of optimization strategies. Traditional focus on Google Search alone is no longer sufficient. Marketers and content creators must understand these AI-driven biases to effectively target and enhance presence across varied platforms.

Conclusion

AI assistants are reshaping digital content landscapes by privileging different sources than conventional search algorithms. Recognizing their unique biases and citation patterns is essential for adapting SEO tactics and ensuring broad visibility in a rapidly evolving search ecosystem.