Podcast
Conductor put fourteen thousand questions to AI assistants. Up to 72 percent of answers to educational questions named no brand, 20 percent of pricing answers named none, and one brand led 91 percent of comparison answers. Adam and Bea go through what each number means for how you measure AI visibility.
Adam: The Content Marketing Institute published a study this month, from Conductor. Fourteen thousand questions put to AI assistants, across ten industries and seven kinds of buyer intent. I read it as the clearest case yet for how we built AI Visibility.
Bea: Bold. I'll be the sceptic, then. Rand Fishkin at SparkToro showed that AI answers change almost every time you ask. So why would anyone pay to track them?
Adam: Hold that question. The study answers it, and it does it with three numbers.
Bea: Go on.
Adam: On educational questions, the "how does this work" kind, up to seventy-two percent of answers named no brand at all. On pricing questions, twenty percent named no brand. And on comparison questions, one brand led ninety-one percent of the answers.
Bea: So the assistant behaves completely differently depending on what the buyer is asking.
Adam: That's the whole point. And it's why one average visibility number tells you very little.
Bea: Start with the seventy-two.
Adam: A buyer asks how to do something in your field. The assistant answers from pages like yours and, most of the time, never says whose they were.
Bea: So my how-to guides are teaching the assistant, and the assistant takes the credit.
Adam: Pretty much. Now imagine those questions sit inside your visibility score. You'd look invisible, and you'd start writing more guides to fix it.
Bea: Which wouldn't fix anything.
Adam: Right. That's why AI Visibility tags every question with a buying stage. Awareness, consideration, decision. The grade gives the most weight to the stages closest to a sale, so a how-to question where nobody gets named counts for much less.
Bea: So I can keep writing guides.
Adam: Keep writing them. People read them, and assistants learn from them. Just don't measure yourself on them.
Bea: Pricing. Twenty percent with no brand.
Adam: That one's the easiest to fix. The price is behind a "contact sales" form, or it's in an image, or a script loads it after the page opens. An assistant that can't read the number can't use it.
Bea: And plenty of B2B companies won't publish a price.
Adam: They don't have to publish the whole price list. Shannon Vize at Conductor said any pricing detail beats none. A starting price and what moves it up is already something an assistant can quote.
Bea: How would I even know I'm losing those questions?
Adam: The Questions screen lists every question we ask for you, and whether you turned up in the answer. A pricing question where nobody gets named is an open seat. You write a plain answer, in text, and the next weekly report asks again.
Bea: Then the ninety-one percent.
Adam: That's the one I care about most. A buyer asks "which tool should I choose?", and the assistant picks one. Being on the list and being the answer are different things.
Bea: And if it picks a competitor?
Adam: Then that competitor wins the comparison in your category, week after week. Nobody clicks anything, so your analytics shows nothing at all.
Bea: Can Adamiro actually see that?
Adam: It's what the report is built around. "Answers that chose" counts how often an answer picked one brand. "Chose you" is your share of those picks. And "Who the answers chose" names the winners, including companies you don't even track.
Bea: That last one sounds uncomfortable.
Adam: It's the most useful thing on that screen. A company you've never heard of winning your category is exactly what you need to know.
Bea: Fine. I know who's winning. Then what?
Adam: You send that question to Voice and pick Decision guide as the structure. Voice drafts it from your own material: the criteria, the trade-offs, and who each option suits, in plain text an assistant can read. Including the cases where the other product is the better fit.
Bea: You'd tell buyers where you lose?
Adam: Yes. A page that only praises you reads like an ad. A fair one is the page a buyer believes.
Bea: Back to my objection, then. The answers still change every time.
Adam: They do, and Fishkin was right about the part that's noisy. He called any tool that sells you a ranking position in AI "full of baloney". What held steady in his data was how often a brand gets named, over many runs.
Bea: And that's what you measure?
Adam: Yes. Answer position is reported, and kept out of the grade on purpose. The grade is built mostly on how often you're named, at each buying stage, every week, across several assistants.
Bea: Why weekly? Why not daily?
Adam: Because the answers move over weeks. A daily report would mostly measure noise.
Bea: And when I publish that comparison page, how do I know it worked?
Adam: The Impact panel asks the same question again after you publish and shows you the before and after.
Bea: So the study you started with…
Adam: …describes the problem AI Visibility was built for. Find the questions where someone else gets chosen, write the page that answers them, and ask again next week.
Bea: And if I just want to see where I stand today?
Adam: There's a free check on adamiro.com. Type your domain, and it asks an AI assistant five questions about your field and tells you how many named you.
Bea: Five questions. That's a start.
Adam: It's a start. The full report is where you win the comparisons.
More episodes