Setting up AI visibility

AI visibility asks a set of language models the questions your buyers ask, and reads each answer for your brand and for a link to your site. It is paid for out of a plan’s monthly AI allowance rather than out of its lookup budget — a separate allowance, which is why a plan either includes AI visibility or does not. Drafting a post and grouping a worklist into topics spend the same allowance.

Prompts are yours to write

A project holds up to 10 prompts, and they should be the questions a buyer actually types rather than the keywords you track. “Best waterproof trail shoes for wide feet” is a prompt; “trail shoes” is a keyword. A project with no prompts is never swept at all, which is how the product makes sure it is measuring something somebody asked for.

What one sweep is

One sweep asks every prompt of every model — 3 of them — so the answers can be compared across labs rather than taken from one. A sweep runs about once a week on its own, and the button starts one immediately. The button has an hour between presses, and the reason is not only the cost: a single check is a sample of a system that does not answer the same way twice, so two sweeps an hour apart measure the same noise twice at full price.

The budget, and what a refused sweep looks like

Each sweep spends the plan’s monthly AI allowance. A workspace whose plan has none is never swept, and a workspace that has spent the month is skipped rather than failed — recording a sweep as failed would look exactly like evidence that no model named you, when the truth is that nobody asked.

What is answering

The thing answering is a model reached through an interface, not the consumer assistant built on top of it. That call carries none of a chat product’s own instructions, memory or retrieval, so a figure here is about the model named beside it and about nothing else. Every model is named by its model name for that reason, never by the name of a product.