AI assistants

Prompts are asked of AI assistants, and each assistant is its own market. Tracking more than one is what turns "our visibility" from a single number into a picture.

Available assistants

Additional assistants may be available on request depending on your plan. Which ones are active for your account is shown in the assistant filter on the dashboard.

Why assistants differ

The same prompt produces genuinely different answers across assistants, because they differ in how much they rely on live sources, how many brands they name, and how they phrase recommendations. Treat a per-assistant difference as information about that assistant, not as an error.

This shows up most clearly in Citations & Sources: assistants that lean on live web sources produce rich citation data, while assistants answering more from internal knowledge cite less. Low citation volume for one assistant is not a data problem.

Choosing what to track

• Start with the assistant your customers use. For most markets that is the default-enabled set.
• Add assistants for coverage, not for the score. More assistants means more answers and a more stable trend, at the cost of more collection allowance per cycle.
• Do not enable one mid-analysis. Adding an assistant changes the answer pool, so the period before and after are not comparable.

Comparing fairly

The dashboard's assistant filter is the tool here. Two rules:

1. Compare one assistant against itself over time. This is the only clean comparison for a trend.
2. Compare assistants against each other only within the same period and group. Different periods or groups make the difference unreadable.

A common finding is strong visibility on one assistant and weak visibility on another for the same prompts. That is usually a source story — the weak assistant relies on sources where you are absent. Filter Sources to that assistant to see which.

Effect on metrics

With no assistant filter applied, metrics pool the answers of every enabled assistant. That pooled number is a fair overall summary, but it hides per-assistant differences — always drill in before acting on a change.

Because each enabled assistant answers each prompt, enabling more assistants multiplies the answers collected per cycle. See Trial and paid plans for how that interacts with your allowance.

Related

• Reporting periods — the other half of scope
• Data freshness — how often each assistant is asked
• How metrics work — what pooling means