← Guides

How AI assistants choose which sources to cite

When an assistant answers “who are the best [category] providers?”, it’s drawing on two things: what it learned in training, and — when it searches — what it retrieves live from the web. Knowing how those work explains why some brands get named and others don’t.

1. Training knowledge

Models absorb patterns from large amounts of text. Brands that are widely and consistently described across the web are more likely to surface from memory. This favors established presence and consistent naming — and it changes slowly.

2. Live retrieval (search-grounded answers)

When an assistant searches the web to answer, it favors sources that are relevant, clearly structured, current, and credible — pages that directly answer the question, are easy to parse, and come from sites it treats as trustworthy. This is the part you can influence fastest.

What makes a brand citable

• Clear, direct answers to the exact questions buyers ask (not buried in marketing copy).
• Well-structured pages (headings, concise definitions, lists) that are easy to quote.
• Freshness — recently updated, verifiable information.
• Corroboration — being described consistently across multiple reputable sources.

Why you have to measure it

Because answers blend memory and live retrieval — and vary run to run — the only way to know if you’re becoming a cited source is to track it over time: presence, share of voice versus competitors, and the monthly trend.