King of AEO

Selection Rate

How often a model picks your brand from the candidate set it could have used when generating an answer. Attributed to Dan Petrovic and DEJAN, 2025 to 2026.

What makes it a better metric

Most visibility metrics count appearances against nothing. Selection rate counts them against the opportunity: of the times you could have been used, how often were you.

That distinguishes two very different failures. Never being in the candidate set is a retrieval problem. Being in it and passed over is a selection problem, and the fixes have nothing in common.

The catch

The denominator is not observable from outside. Nobody publishes the candidate set, so any selection rate is estimated by proxy: sampling prompts, inferring who could plausibly have been used, and comparing.

The concept is sound and the measurement is approximate. Tools reporting selection rate to a decimal place are reporting their model of it, not the thing itself.

Related terms

Sources

  1. DEJAN AI
  2. AI SEO Wiki — Selection Rate