Why source-selection rate should replace clicks as your GEO KPI

Most GEO dashboards I audit still treat clicks as the headline success metric, which is a category error. In a generative answer the user may never leave the surface, yet the citation still did commercial work. The measurable moment is not the click — it is the retrieval decision: whether your domain was selected as a source, and how consistently that happens across prompt variants, engines and intents.

Source-selection rate is straightforward to operationalize. Define a fixed prompt set (30–50 queries per intent cluster), run it across the major engines on a weekly cadence, and log for each run whether your domain appears in the cited set, at what position, and alongside which competitors. Selection rate = cited runs ÷ total runs. Segment by intent, engine and content type, and you get a defensible trend line for a board deck rather than a screenshot folder.

The harder discipline is pairing volume with quality: a citation inside a hallucinated or off-brand answer is not a win. We weight selection rate by answer accuracy and brand framing before it reaches any executive summary, using the [AI Engine Optimization framework](/framework/aieo) as the scoring spine.

If your leadership still asks for click-through on AI surfaces, what evidence have you used to reframe the conversation?