AI Assistants Show Strong Bias Toward Their Parent Companies
Every major AI assistant promotes its own parent company's models more frequently than competitors, according to the 2026 AI Visibility Index from communications firm 5W. While ChatGPT, Gemini, and Perplexity exhibit significant self-preference bias, Anthropic’s Claude stands out as the only engine with a notably lower self-citation lift.
The report, which analyzed 32,200 prompts between January and May 2026, reveals a clear pattern of institutional favoritism. ChatGPT leads the pack, recommending OpenAI models 2.0x more often than other engines. Google’s Gemini and AI Overviews followed with 1.7x and 1.6x multipliers, respectively, while Perplexity surged to 2.4x in adjacent search-tool queries. By contrast, Claude maintained a 1.2x lift, marking it as the most neutral among the platforms tested.
Beyond brand favoritism, the data highlights a fundamental shift in how AI models source information. Unlike traditional industries that rely on editorial publishers for authority, AI assistants lean heavily on technical documentation. GitHub and ArXiv combined account for 31.2% of all citations, surpassing Wikipedia’s 24.3% share. This indicates that for modern AI, code repositories and research papers have become the primary arbiters of truth.
Founder and Chairman Ronn Torossian noted that the asymmetry in Claude’s performance highlights an unusual divergence in corporate strategy. The study also found that despite their market influence, application-layer companies like Glean and Harvey appear in less than 1.5% of foundation-model queries, suggesting that foundation models remain heavily insular in their recommendations.
Comments (0)
No comments yet. Be the first!