How can we read AI visibility metrics without falling for inflated numbers?

The number nobody had a year ago

Every B2B marketing team suddenly has a number for how often AI recommends them. A year ago that number didn't exist. Now there's a dashboard for it. A share-of-voice percentage. A little green arrow. A slide in the board deck that says AI visibility is up and to the right.

Here's the uncomfortable part. A lot of those numbers are measuring the wrong surface, seeded with the wrong prompts, and filtered to look better than the truth. You can hand yourself a victory in AI visibility that you never actually won. Plenty of teams already have, and they're reporting it up the chain with a straight face.

Does that mean the tools are worthless? No. It means a dashboard is only as honest as the questions behind it, and almost nobody is checking the questions. The score on the screen and the truth about whether ChatGPT recommends you when a real buyer asks are two different things. This is just truth.

What's broken here isn't the technology. It's the metric. We've taken the vanity number that hollowed out the last decade of marketing, the impression, dressed it up in AI clothes, and started celebrating it all over again.

What is a vanity AI-visibility metric?

A vanity AI-visibility metric is any 'how often does AI mention us' score that looks impressive because of how it was measured, not because of what's happening in the answers real buyers see. It inflates through three moves. Once you know them, you can't unsee them.

1. 1Brand-seeded prompts. The tool asks the model questions that already contain your name. 'What makes Acme a good messaging platform?' You'll show up almost every time. Of course you will. You handed the model the answer. That's measuring recall, not discovery.
2. 2Biased query sets. The prompts skew toward the categories you already win. Run enough questions you're strong on, quietly skip the ones you're weak on, and the average climbs. The trouble is your buyer never saw your query set.
3. 3Selective filtering. The denominator shrinks. Prompts where you don't appear get tagged 'not relevant' and drop out of the count. The percentage rises because the losses left the room.

None of this needs a shady vendor. Most of it happens by accident, because the person building the prompt list wants to see the brand do well, and flattering defaults are easy to reach for. The result is the same either way. A green number that doesn't survive contact with a real buyer's question. If you want the honest version of the work, how do you measure whether AI engines are recommending your B2B company walks the clean method.

Why is this getting worse in 2026, not better?

AI brought the cost of content to zero. We watched a whole industry answer that by making more of it, faster, and calling the volume a win. Now the same reflex, do more and prove it fast, is coming for measurement. There's a cottage industry selling AI-visibility scores, and the incentive runs one way: sell a number that goes up, not a number that's true.

The deeper problem is that most dashboards blend two completely different things into one percentage. There are two surfaces where AI can recommend you, and they behave nothing alike.

This is why brand is the new backlink. In AI search, a clear and consistent brand narrative is what gets you cited, the way backlinks once drove search rankings. A dashboard that mashes those two surfaces into a single number is like a bank statement that adds your checking balance to your thirty-year mortgage payoff and hands you one figure. Technically math. Completely useless. When you can move one surface in a week and the other takes a year, you have to track them apart, or you'll credit a blog post for a shift it could never have caused.

And the gap between ranking and getting recommended is real. Walker Sands, studying 828 enterprise B2B companies across more than 45 million queries, found the median company gets cited in just 3 percent of AI Overviews even on keywords where it already ranks number one. Showing up in the crawler's index is not the same as showing up in the answer, which is the whole point of why doesn't AI cite my B2B company when buyers ask for recommendations and, more bluntly, why does my B2B company rank on Google but not get cited in AI search.

How do you pressure-test an AI-visibility number? Seven questions.

You don't need a tool to check a tool. You need seven questions and an honest afternoon. Run your dashboard through these before you put it on a slide.

1. 1Which surface is this measuring, grounded or parametric? If the tool can't tell you, the number is noise. A blended score hides which lever actually moved.
2. 2Who wrote the prompts, you or the buyer? If the questions contain your brand name, you're measuring whether the model remembers you, not whether it chooses you.
3. 3Are branded and unbranded queries separated? Winning 'is Acme a good messaging firm' is not winning 'who's the best messaging firm.' Only the second one is a sale.
4. 4Is the query set how your buyer actually asks? Pull ten questions from real sales calls and run those. If your score craters, your dashboard was curated.
5. 5What's the denominator, and is it stable? If prompts you lose get dropped as 'not relevant,' the percentage is inflated by subtraction.
6. 6Does the number move when you changed nothing? Ask the same model the same question three days running. If the answer swings, some of your 'trend' is just model variance.
7. 7Can you read the receipts? If you can't see the actual answers and the actual sources the model cited, you're trusting a score you can't audit. Read five real responses by hand.

You can run the whole thing in an afternoon. Open the model your buyers actually use, ask it the questions they actually ask, in a clean session with no brand name in the prompt, and read what comes back. That five-minute read is more honest than most dashboards, and it costs nothing. If you want the map of how AEO, GEO, and SEO measure differently in the first place, what's the difference between AEO, GEO, and SEO for B2B founders lays it out.

What do we see across 100+ B2B companies tracking this?

The pattern is almost boring. When we re-run a company's AI-visibility number with clean, unbranded, buyer-real prompts, the honest score is almost always lower than the dashboard, sometimes by half. The teams most convinced they're winning AI search are usually the ones measuring brand-seeded recall and calling it discovery.

The second pattern is surface confusion. A team ships a strong owned-content push, watches grounded visibility climb, and reports 'we're winning AI.' Meanwhile parametric visibility, the memory baked into the model, hasn't moved a point, because no third party changed how they describe them. They're winning the surface that moves in weeks and losing the one that decides the hard recommendations. Both are real. Only one of them made the slide.

A real example: our own scoreboard, including the ugly number

We'll use ourselves, because it's the cleanest way to show what honest looks like. PitchKitchen tracks its own AI visibility on both surfaces and reports both out loud, including the one that stings.

Over about six weeks, our grounded visibility, the surface the model retrieves live, moved from roughly 12 percent to about 19 percent. Real progress, driven by owned content that finally tells one consistent story. Our parametric visibility, the memory baked into the model from what the rest of the web says about us, is still close to zero. We say that in the same breath as the good number.

Why volunteer the ugly half? Because the day we blur those two into one flattering percentage, we've started lying to ourselves. A scoreboard that can't show you losing isn't a scoreboard. It's a mirror you've decided to like. The honest read told us exactly where the next quarter of work goes. Not more blog posts. Getting independent sources to describe us the same way our own pages already do.

What this means for you

If you take one thing from this, take this. Separate the two surfaces, and make your tool show you the losing number. The moment you do, the wasted tactics fall away on their own.

1. 1Split every AI-visibility report into grounded and parametric, and refuse to look at a blended number. They move on different clocks and need different work.
2. 2Rebuild your prompt set from real, unbranded buyer questions, and read the actual answers by hand once a month. Trust the receipts over the score.
3. 3Stop buying tactics that can't move the surface you're losing. Paid mentions and content volume don't change parametric memory. Independent consensus does.

Here's the part the dashboard can't tell you. The thing that moves both surfaces, the only thing that gets the whole web and every AI model to describe you the same way, is a clear narrative identity, documented once so it repeats everywhere without drifting. That's what a Magnetic Messaging Framework is. The documented brand bible that gives your own content, your buyers, and the model one true story to converge on, which is also how you get your B2B brand to show up in ChatGPT and Claude recommendations in the first place. The measurement tells you whether the consensus is forming. The narrative is what forms it. Fix that, and the numbers stop needing to be gamed, because there's finally something real to count.