AI visibility vs AI revenue attribution
These get sold as the same thing and they are not. One measures whether an AI engine names you when a buyer asks. The other measures what happened to the people who arrived because it did. They break independently, they need different fixes, and confusing them is how founders spend a quarter on the wrong problem.
Side by side
The four states you can be in
Because the two measurements are independent, there are four combinations, and each one calls for completely different work. Getting the diagnosis right is worth more than any individual tactic.
Invisible and no traffic. You are not named in answers and almost no AI traffic arrives. Check crawler access before anything else. If GPTBot and PerplexityBot have never fetched the pages that would answer your buyers' questions, the problem is robots.txt, server-side rendering, or a signup wall, not your writing.
Visible but no traffic. Engines name you and almost nobody arrives. This is partly real and partly an illusion. Being mentioned in an answer does not guarantee a click, and many of the clicks you do get arrive with the referrer stripped, so they land in your direct bucket. Before concluding that citations are worthless, check whether direct traffic rose alongside your visibility trend.
Visible with traffic that does not convert. The engines send people and the funnel loses them. This is the most tractable state, because you have live sessions to inspect. Look at where the AI-referred cohort drops out compared with your other sources. AI-referred visitors often arrive with an expectation set by the answer they read, and a mismatch between that expectation and your landing page is a common cause.
Visible, converting, and growing. Now the work is defending the position. Track share of voice against named competitors, watch for sentiment slipping from recommended to merely mentioned, and keep the pages that earn citations accurate and current.
Why one number for both is a bad idea
It is tempting to combine mentions, traffic, and revenue into a single AI score. Resist it. The inputs have different reliability: a citation is a non-deterministic model output, a session is an observed fact, and a payment is a confirmed fact from a provider webhook. Averaging a guess with a fact produces a guess wearing a fact's clothing.
Worse, a composite hides the diagnosis. A score that drops could mean you lost citations, or that traffic fell, or that conversion dropped, or that a payment integration broke. Those need four different responses, and a single number tells you none of them. Metrivo keeps crawl access, citations, sentiment, AI sessions, and attributed revenue as separate readings for that reason.
The same logic applies to converting citations into estimated revenue. There is no defensible multiplier from "named in an answer" to "dollars." Click-through from generated answers varies enormously by question type and answer format, and nobody publishes reliable figures for your category. A model that invents one will be confidently wrong.
What most tools actually cover
The market splits cleanly. AI visibility monitors run prompts across engines and report mentions, sentiment, and share of voice. They are good at that and structurally blind to your funnel, because they have no tracking on your site and no connection to your payment provider. If you buy one, you will learn whether you are recommended and nothing about whether it pays.
Analytics and attribution tools work from the opposite end. They see sessions, referrers, and payments, so they can tell you what AI-referred visitors did, but they never ask an engine anything. If you buy one of those alone, you will optimise the conversion of whatever traffic you happen to get and never learn that a competitor is being recommended ahead of you.
Metrivo runs both halves and keeps them labelled. Read the AI Visibility page for how the prompt scans and crawler log work, and AI Search Attribution for how AI sessions are joined to payment evidence. The detection limits that apply to both are written out in AI traffic detection.
