TL;DR

Your AI visibility dashboard may say you’re winning when you’re not even in the room. Many vendor tools test prompts that paraphrase your own content, so of course you show up. Ask the questions real prospects and customers type, and you may find AI handing the answer to government agencies, aggregators, and competitors with original research. The fix is an honest AI visibility audit, an anchor hub, a trust map built from what AI actually cites, and a measurement plan that answers “says who?” None of it is hard. It just isn’t anybody’s job, and that’s why most teams never finish.

Key Insights

• AI assistants are already answering your customers’ questions and recommending three or four options at most. If you’re not one of them, you’re invisible at the moment of decision.
• The AI tools cite the same source only 1.7% of the time, so winning in one assistant doesn’t mean you’re showing up in the others.
• Many AI visibility dashboards test paraphrases of your own content. Of course you show up. It’s like Googling yourself and being impressed you’re on page one.
• An honest audit starts with the 20 questions customers actually ask, tested across five AI tools, with every answer logged as absent, mentioned, cited, or featured.
• The sources AI cites for your questions become your trust map: journalists, bloggers, podcast hosts, newsletter authors, and anyone else with the audience you want.
• If you can’t answer “says who?” about a metric, it doesn’t go in front of executives. That includes the good news.
• Teams rarely fail because the work is hard. They fail because of discipline: the prompts get easier, the Winnable column becomes a wish list, and the anchor hub goes stale.

Your AI Visibility Dashboard Is Lying to You 

Do me a favor. Open ChatGPT (or your AI assistant of choice) in an incognito window and type the question you think your best customers are asking to find your organization (or worse, your competitors).

Not your headline. Not your tagline. Not your key messages. Not what you think people use when they’re searching for a brand like yours.

Their actual words—the clunky, plain-language version they’d type or speak out when they’re trying to figure out what to do. This kind (which I actually used for my sister-in-law the other day): “I accidentally dropped the lid of my laundry detergent down the lint trap in my dryer. What do I do?”

She really did do that, and it turns out, leaving it in there is a massive fire hazard!

Now it’s your turn. I’ll wait. Go ahead and type in what you think people are using when they’re searching for products or services or organizations like yours.

Okay. Who did it name?

Now do the same thing in temporary mode in Claude. And Gemini. And Perplexity. And Copilot. You’ll probably get five different answers.

In their most recent blogger survey, Orbit Media found that the AI tools cite the same source only 1.7% of the time.

So if you were hoping to win one and call it a day, think again. They’re the same less than 2% of the time.

Here’s the thing. That answer is being delivered to your prospects and customers every single day, whether you’ve looked or not.

Most teams haven’t looked.

And the ones who have? They’re often looking at the wrong thing.

What Happens When Your AI Dashboard Says You’re Fine

Let me tell you about a client of ours. We recently spent some money on an AI visibility test and got some pretty interesting early results that led their team and mine to form a hypothesis we wanted to test.

Initially, the vendor report on their GEO visibility was incredibly strong. It showed they were beating most of their competitors across everything that mattered to them, including in places where they wouldn’t normally be able to compete, and against the really big guys.

That’s what initially made me wonder whether the data was accurate, so we started digging into it. We started with where they were showing up, and for which prompts. When we tried to manually replicate the results on five AI assistants, we weren’t able to.

So then we dug into the prompts the vendor was using and found they were paraphrases of the client content.

Of course they showed up for that! They were asking AI about their own content and, because they were the only ones talking about it, they were mentioned, cited, and featured.

That’s like Googling yourself and being impressed you’re on page one. Good Thor.

So we swapped in the questions human beings type into ChatGPT, and the client was absent from every single answer.

AI handed those questions to government agencies, aggregators, and online tools. And when a company in their space did get named, it was a competitor—one whose original research anchored the answer.

They weren’t being described incorrectly. They simply weren’t in the room.

Phew.

The good news is that we now know and have a solid baseline for our monthly measurement reporting. AND, within the first month of measurement, their newly consolidated how-to content was already appearing in Google’s AI search surfaces—the first confirmation that when movement happens, the instruments will see it.

That’s a first reading, not a trend.

It’s not the end of the story, either. It’s not even the middle. We’re at the very beginning, and we have enough data to know where they stand, where the gaps are, and what we need to add.

The Answer Is Being Written Without You

Here’s why this matters right this very second and can’t wait for you to someday get to it.

Your buyers are building shortlists before they ever talk to sales—and it doesn’t matter if you’re in B2B, B2C, D2C, non-profit, ecommerce, SaaS, education, or anything in between.

Every one of us is making decisions based on what our AI assistants tell us. And unlike a Google search, where we got 10 options and could do our own research, our AI assistants give us only three or four options at best and typically make recommendations.

I’ve already made the case that AI agents are the new gatekeepers, so I won’t rehash it here, but we have to figure out pretty quickly how to be the ones standing at the front door.

To do that, your content has to be recent. This isn’t a “set it and forget it” job where you optimize one page once and coast. The brands that start feeding the system now are the ones it learns to trust.

The good news? The window is still open. Almost nobody owns this yet—only 3% of PR pros say they do. That won’t last. The competitors who start this quarter will be the ones AI quotes next year.

You don’t need to win the AI answer. You need to stop being absent from it. It takes some time, but starting now is the best idea you’ll have all quarter.

The AI Visibility To-Do List

There are four things you can do right now to execute on that idea—and ensure you’ll show up in the AI answers.

1. Audit what AI says about you now (and set an honest baseline)

Start by listing the 20 questions customers actually ask, in their own words. Pull them from sales calls, customer service tickets, and search data—not your editorial calendar.

Ask all five AI tools the 20 questions in incognito or temporary mode. That’s 100 prompts.

For each one, log if you are absent, mentioned, cited, or featured. Note which competitor gets named and which sources AI cites. You’ll need those sources in step three.

In our spreadsheet, the column headers are: Prompt, Appearance, Winnable, What We Want Included, and then each AI assistant has a column.

Here’s what goes in each column:

• For the Prompt column, we paste in the actual questions people use when looking for a client’s (or our own) brand.
• For the Appearance column, we want to know if they’re absent, mentioned, cited, or featured. That’s just a simple dropdown.
• For the Winnable column, we’re looking at who is mentioned, cited, and featured—and for what (original research, expertise we don’t have)—and making a judgment call. Do we think we can win for that prompt, or not?
• For the What We Want Included column, this is more art than science. We’re including what we’d love to have included in the AI response in the future. This is the start to our plan.
• And then, in each AI assistant column, we list who appears in the full answer, not just competitors. This is because most of the time ancillary businesses show up, such as government agencies, aggregators, and online tools we found for our client.

Then do the consistency check. Pull your website, news releases, social posts, and media coverage, and look for every place you describe yourself differently.

When your PR, social, and sales teams say three different things, AI either makes something up or ignores you. You have to, have to, have to be consistent.

2. Outline your first anchor hub

Next, you’re going to outline your anchor hub. If you don’t know what that is, why it’s important, or what to include, I highly recommend you watch the replay of our Visibility Engineering™ webinar. It’s free and available on demand, so you can set it to 1.5x or 2x speed and get caught up.

Once you’re ready, choose the one customer question that matters to your organization the most right now. For us, that might be, “How do I use the PESO Model® in 2026?”

The most effective anchor hubs do five things:

1. Frame the page as the buyer’s question.
2. Define the problem clearly.
3. Name your method or framework, so there’s something specific to cite.
4. Show evidence and outcomes: data, results, case studies, testimonials.
5. Include quotes and data that AI can safely cite word-for-word.

Take a look at an existing page on your website that loosely answers the one customer question you identified and score it against those five things. If it doesn’t do all five, you have your homework cut out for you. When you can say, “Yes, it does this” to all five, you have a strong anchor hub.

3. Build a media list around what AI actually reads

Now it’s time to build your media list around what AI reads. Actually, we stopped calling it a media list. We now call it a trust map because it includes more than media. It includes anyone who has an audience you’re trying to reach—bloggers, influencers, newsletter authors, podcast hosts, and yes, journalists, too.

Go back to the work you did in step one. Look at the spreadsheet and note who is mentioned, cited, or featured for each prompt (other than you, if you’re in there). This is the start to your trust map. If they’re already being cited, pitching them in whichever way is appropriate is the first step.

Then map each partner to a topic your experts can credibly own. Keep in mind, though, this only works if your expert has an actual point of view, which is the whole argument in last week’s thought leadership article.

4. Build a measurement plan leadership will believe

And, then of course, we need to build a measurement plan because it wouldn’t be content with Gini without measuring our work. You’re welcome!

For every metric you use, you have to be able to answer “says who?” If a number is modeled, estimated, or calculated by a vendor formula, it doesn’t go in front of executives until you can corroborate their claims.

A year or so ago, I was in a meeting with a client’s data and analytics team and I asked for some very specific marketing data. The head of the team looked me square in the eye and said no.

I thought he was joking! He was not joking. When I asked why, he said, “Because marketers like to take the data and use what makes them look good and ignore the rest. I believe strongly in not allowing that to happen.”

Bruh. Me, too.

I finally talked him into giving us access when I showed him that we were corroborating data one of their vendors had sent—and I had a hypothesis that it was wrong. He and I dug into it and lo and behold! I was right.

It was instant trust, and I got the data we needed to be able to make decisions.

You can’t report solely on the things that make you look good. You have to show the down numbers too, because that helps you make really smart decisions. It also helps you figure out how to best show up in AI visibility—and how to measure any movement.

Why Most Teams Never Finish

As you can see, this isn’t very hard. The challenge is that it’s not really anybody’s job, so it gets sidelined until an exec or client asks for data, and then everyone scrambles to pull data that tells a glowing story. And sadly, that data is often from an AI visibility software that only shows you if you show up for your own content.

Wah. Wah.

Take ownership of it. Build your spreadsheet. Make it part of your process.

What Usually Goes Wrong When Teams Do This Alone

Here’s what we see most often. The spreadsheet gets built. Everyone feels great about it. And then one of four things happens.

The prompts get easier. Somewhere around prompt 14, someone runs out of real customer questions and starts writing the ones they know they’ll win. That’s how our client ended up with a dashboard full of good news. Nobody did it on purpose. It’s just human nature to want to see your name in the answer.

The Winnable column turns into a wish list. It’s supposed to be a judgment call—who’s cited, why, and whether you have the evidence to compete. Instead, every row says “yes” because no one wants to be the person who says you can’t win. 

(Someone has to be that person. Be that person. Please be that person!)

The anchor hub launches and then just sits there. Remember, AI favors recent sources. One great page in October is a lovely page in October. By March, it’s stale and old and all of that work you did was for naught.

And the trust map quietly turns back into a media list. The bloggers, podcast hosts, and newsletter authors fall off because they’re harder to pitch than the trade reporter you’ve known for 10 years.

None of those are skill problems. They’re discipline problems—and discipline is the hardest thing to install from the inside, when everyone is already doing three jobs.

You Might Be Here If…

• You’ve never typed your top customer question into an AI tool.
• You did, saw a competitor’s name, and closed the tab.
• Your AI visibility report looks great, and you’re not sure what questions it asked.
• Two teams have two dashboards with two different answers.
• “AI visibility” has been on the agenda for two meetings and nobody owns it.
• You just don’t have time to focus on it, even though you know how important it is.

The Test Is Still Waiting

Let’s go back to my sister-in-law and her dryer.

When she typed that question, she didn’t want 10 blue links. She wanted someone to tell her what to do before her laundry room caught fire. AI gave her an answer and a source, and she trusted both, got the lid out of her dryer, and her clothes no longer have to air dry.

Somebody earned that spot. Not with the best tagline or the biggest ad budget, but by answering a real question, clearly, in a place AI trusts.

That’s the whole game. Not the dashboard. Not the vendor report. The question your customer actually asks, in their words—and whether you’re in the answer.

Our client thought they were. It took one honest test to learn they weren’t even in the room, and it’ll take months of disciplined measurement before anyone gets to say they are. 

I’ll take honest and slow over flattering and wrong every single time.

So go back to that incognito window. Ask the question. Write down what you see, even if it’s nothing.

Especially if it’s nothing.

And if you’ve built the spreadsheet and your Winnable column is mostly “no”—or your vendor dashboard and your manual tests don’t match—that’s exactly the work we’re doing for the client in this article. 

Tell us what you found, and we’ll show you how we’d build your baseline and the plan to get you in the room.

© 2026 Spin Sucks. All rights reserved. The PESO Model is a registered trademark of Spin Sucks.