1,790 AI search measurements reveal why most visibility scores are misleading

Short summary

The author ran 1,790 measurements across five AI engines and found that most 'AI visibility' dashboards are artifacts of branded-prompt averaging — one company's 35% score dropped to 0% when brand-name prompts were removed. Different engines have different citation supply chains, so a single visibility number is meaningless without specifying engine and prompt set. The article identifies three mirages in AI visibility reporting and proposes four numbers worth tracking instead.

• •Branded-prompt averaging inflates AI visibility scores — one company's 35% became 0% on organic-only prompts
• •Different AI engines cite different sources: Perplexity favors Reddit, AI Overviews inherit Google rankings, ChatGPT sits in between
• •Single-number 'AI visibility scores' are marketing unless they specify engine, prompt set, and branded vs organic separation

Generated with AI, which can make mistakes.