AI visibility gets funded when somebody can point at pipeline. The hard part is that the channel hides itself: a buyer asks an assistant, reads an answer, and arrives later as direct or branded search.
Attribution you can defend
Assume you will not get clean last-click attribution and stop trying. What holds up in a budget conversation is a chain of evidence: the prompts your buyers ask, whether you are recommended in them, and what happens to branded demand and direct traffic as that recommendation rate moves.
Self-reported attribution on the demo form is the single cheapest thing you can add. One open question, "how did you hear about us", catches what analytics cannot.
The number to lead with
Lead with recommendation rate on purchase-intent prompts, and show it beside pipeline from branded and direct. Those two moving together is the argument. A chart of citations alone is not, because nobody buys a citation.
Table stakes
- A split between research prompts and buying prompts, measured apart.
- Self-reported attribution on every inbound form.
- Branded search and direct traffic tracked as the downstream signal.
- A cost per probe you can state, so spend is a known number.
- One quarterly view that puts recommendation rate next to pipeline.
Be careful what you promise. This channel compounds slowly and then quickly, and a forecast built on a straight line will embarrass you.