Is an AI visibility dashboard enough?

For most small marketing teams, no. A dashboard counts how often ChatGPT, Gemini and the rest name you for a set of questions and draws that as a line. It can't tell you why the line moved, whether the questions were fair, or which page to change. It's enough when you already have people who know what to do with a falling line.

If you're paying for one and the line just dropped, that gap is probably why you're here.

What does an AI visibility dashboard measure?

How often you turn up in answers to the questions you gave it. You supply prompts, it asks them on a schedule and reports the share of answers that mention you. Most also show which competitors appeared alongside you.

When it drops, the chart can't say which of your pages an assistant read, what it found there, or what would have changed its answer.

Why does the number change when nothing on my site did?

Assistants rarely give the same answer twice. SparkToro had volunteers run the same prompts through ChatGPT and Google's AI nearly 3,000 times, and the same list of brands came back in fewer than one run in a hundred. Model updates, whatever the assistant retrieved that day and plain chance all move the line.

Read it the way you'd read a poll. One day's figure says little. A share across many questions, compared with the same questions a month earlier, says something.

Does it matter who writes the questions?

It decides more of the result than anything else on the dashboard. Write the prompts yourself and you'll phrase them the way you describe your own business, so the reading comes back flattering. A buyer who hasn't heard of you can't ask that way. They ask about their problem or about the category, and that's where you most need to show up.

Fair questions mostly don't name you. They describe a problem, compare kinds of solution, or ask what to look for. We got the shape of them wrong at first. Our first question writer made every question two sentences in the first person, with an admission of what the buyer wasn't sure about, because that's how a lot of people talk to a chatbot:

"There's no one here whose actual job this is, so it lands on whoever has capacity and it's usually the people we can least afford to lose for a week. How do other businesses handle this without a dedicated team?"

Then we checked it against real questions. In a public set of ChatGPT conversations we found 839 about buying a product or service. The middle one ran to 24 words, and only one in four used the first person at all. Plenty were typed like a search, closer to "payroll software for small business". Those matter more than their length suggests: Peec AI found keyword-style prompts surface about 25% more brands than conversational ones. So a fair set asks both ways, and only the last few questions name you, asked the way a buyer asks once they've found you and are weighing you up.

That's how we build a question bank in Vizebel. At most a quarter of it names you, all at that last stage. The rest name nobody, only the category, or one competitor. Google's AI Overview always gets the short, search-shaped version, since that's what people type there. You can edit or remove any question before it's asked.

What should I use alongside the number?

You need something that tells you what to change: which page an assistant would read for a question, and what's wrong with that page in words you can act on.

Our page review marks a page out of 100: 70 points for the writing and 30 for whether a crawler can reach and read it. The review says where each point went. Does the page open with a clear answer? Do the headings say what each section answers? Does it show first-hand experience? Does it read as if a person wrote it? A page that spends three sentences clearing its throat before it answers anything is something you can fix this afternoon.

We're new, so we can't show you a customer's before and after yet.

Then measure. Save the questions that matter as a snapshot, change the page, and ask the same questions again once the change is live.

When is a dashboard enough on its own?

When you already have the people. A business with an in-house content team, an SEO lead who knows what makes a page worth quoting and a developer who acts on crawler reports doesn't need walking through the work. It only needs the measurement, so it should buy the best dashboard it can afford.

Most small marketing teams aren't set up like that. For them, "down four points this week" leaves the weekly meeting with nothing to do.