Before an AI engine answers, some of them search first. Those background queries are a rare, direct look at how an engine actually researches your category, when the engine is willing to show its work.
Gemini, Claude and ChatGPT hand back the real background queries. The other two get an honest label instead of a guess.
Every query you see is one an engine actually ran and disclosed. If it isn't real, we don't show it.
Fan-outs are captured on every scheduled run, so the query mix stays current as your prompts do.
Some tools show a single fan-out count that blends real, provider-disclosed queries with ones reconstructed or inferred from the answer text. llemmy doesn't blend. We show only what an engine actually hands back, and we label the rest plainly instead of guessing.
Every disclosed query gets scanned for brand mentions, yours and your rivals'. See how often a brand appears inside the background searches themselves, not just the final answer, with the sample size and confidence interval behind the number.
Fan-outs are not a vanity count. We show exactly what an AI engine discloses and label the rest instead of inventing it: real and provider-disclosed, count-only, or not exposed. Brand-in-fan-out proportions carry their sample size and a 95% confidence interval. Measured over the official model APIs and Google AI Overviews read off the SERP, not the logged-in apps; personalization isn't modeled. Full limits in our methodology.
The mention rate, with confidence intervals, that a tracked fan-out feeds.
Share of voice and threat tracking across every rival, on every engine.
First insights in under five minutes. No credit card. Every query we show you is one an engine actually disclosed.