See who AI recommends instead of you, and how often

The proportion of relevant AI answers in your category that recommend you rather than a competitor, measured on unbranded prompts where the buyer has not named anyone yet. It is computed from the same stored answers as everything else, so any share figure can be opened up into the specific responses behind it. Branded prompts are excluded by design: an answer to a question that already names you is not evidence of category standing.

Five on Starter, ten on Professional, and a volume set per deal on Enterprise. A competitor slot covers that brand everywhere it is measured, share of voice, head-to-head, citations and off-site presence, rather than being metered per report. You can swap a competitor out as your market shifts, and the data already captured stays attached to the answers it came from, so a swap does not erase what you have measured.

Yes. Head-to-head views take multiple competitors, so you can see the whole field on a prompt rather than a series of one-against-one charts. That matters because AI answers usually list several options together, and the useful question is often who else made the list, not just whether one named rival beat you. You can filter the comparison by tag or by engine to isolate a product line or a single surface.

On those questions, yes, whatever your positioning deck says. Engines answer from what has been written about your category, so review sites, marketplaces and adjacent products often turn up in recommendations. We suggest competitors from what actually appears in your captured answers, and you decide which to promote into tracked slots. An unexpected name appearing repeatedly is usually a content or citation gap rather than a data error.

Yes. Share of voice deliberately focuses on unbranded, category-level prompts, but you can track branded and comparison prompts too, the 'X vs Y' and 'alternatives to X' questions, and see how each engine handles them. Those are where a competitor's comparison page often does the talking for them, so it is worth knowing which sources an engine leans on. The two views answer different questions: one measures reach, the other measures framing.

Slowly, and that is the honest answer. It is a frequency measurement across repeated daily runs, so it settles over weeks rather than responding to a single publish. Day-to-day wobble on one prompt is normal variance in how engines answer, not a rival's campaign. Wait for a shift that holds across several prompts and more than one engine before treating it as real, and expect earned citations to move it before on-site changes do.