Here’s a way to lose money without ever noticing: get cited by ChatGPT on fifty different topics and get recommended on none of them. Being a source and being the answer are not the same job. And most brands chasing AI visibility are optimizing for the wrong one.
Hi, I’m Jeff Payne. You’re listening to The Jeff Payne Show, Episode #49: Cited Everywhere, Trusted Nowhere.
Quick vocabulary, because this only works if we’re speaking the same language.
When you ask ChatGPT something, it does two separate things.
First, it pulls in sources — footnotes, basically. These are sites it leans on for facts. That’s a citation.
Second, sometimes — not always — it actually names a brand in the answer itself. For example, “You should look at (fill in the blank).” That’s a mention. Being the footnote and being the headline are very different jobs.
And a category, for our purposes, is just a customer question — or really, a cluster of ways people ask the same question. “Best CD rates for teachers in Texas” is a category. So is “How is CD interest taxed?” Every business sits inside dozens of these.
Now — once you’ve become the headline in one category, the obvious next move is: can I be the headline somewhere else? New market, new service line, adjacent topic. Kevin Indig at Growth Memo just ran the numbers on exactly that question, using SEMrush data across a thousand categories. And the answer is not what most people assume.
Getting cited outside your lane is easy. Getting recommended outside your lane is not.
In categories far from a brand’s core expertise, half the time they still show up as a footnote — a source. But they only get named as the actual answer about a quarter of the time. Compare that to categories close to what they’re known for: cited three-quarters of the time and named as the answer 44% of the time.
So the footnote job travels fine. The headline job barely leaves the building.
Here’s where it gets useful. The instinct is “don’t spread yourself thin.” But the data says something more precise: it’s not breadth that hurts you. It’s shallow breadth.
Show up in just one out of five ways a question gets asked in a new category, and your odds of being named actually go down — measurably negative. Show up in all five — meaning you’ve actually gone deep, not just dabbled — and that penalty flips positive. Being everywhere doesn’t cost you. Being everywhere thin does.
And here’s the number that should change how you think about expansion: when a brand gets cited in a category closely related to what it already does, it converts that citation into an actual named recommendation 46% of the time. In a distant, unrelated category? 18%. Less than half.
Translation: Your existing authority is a bridge. It reaches next door. It does not reach across town.
Before you expand into anything — a new topic, a new service, a new market — ask which job you’re actually trying to win. Do you just want to be a credible source people can find? Fine, breadth is cheap, go ahead. But if you want to be the recommendation — the name AI says out loud — you need the closely-related categories, and you need to go deep in each one, not skim five of them.
One field of expertise, tested rigorously in the direction it’s already pointing, beats five directions tried once. This isn’t a new rule. It’s the same one — proof beats proximity — just showing up at the border, deciding whether your authority gets a passport or gets turned away.
CITED EVERYWHERE, TRUSTED NOWHERE.
What new AI-search data says about expanding your brand into new categories — and why being a source everywhere doesn’t make you the answer anywhere
Once a brand earns a strong position in one topic, the next question is almost automatic: where else can we win? New research from Kevin Indig at Growth Memo, drawing on SEMrush AI Visibility Toolkit data across 1,094 categories, tested exactly that — and the results complicate the usual advice to “stay focused” or “go broad.”
TWO DIFFERENT JOBS, ONE CONFUSING WORD
When an AI assistant answers a question, it can do two distinct things: cite a source or name a brand as the recommendation. A citation is a footnote — evidence the AI leaned on to build its answer. A mention is the AI actually naming a brand as the thing to consider.
Businesses tracking “AI visibility” often conflate the two, but the data shows they behave completely differently once a brand steps outside its core expertise.
Being the footnote and being the headline are very different jobs.
CITATIONS TRAVEL.
RECOMMENDATIONS DON’T.
In categories distant from a brand’s demonstrated expertise, the brand still gets cited as a source about half the time — but only gets named as the recommended answer roughly a quarter of the time.
In categories closely related to what the brand is known for, both numbers jump: cited about three-quarters of the time and named outright in 44% of answers.
You can be a source almost anywhere. You’re only the answer where you’ve earned it.
The real penalty isn’t breadth.
It’s shallow breadth.
The instinct to avoid “spreading thin” turns out to be slightly wrong. Brands that appear in only 1 of 5 prompt variants in a new category see a measurable drop in mention share. But brands that show up in all five — genuinely deep presence, not a single appearance — see that penalty flip positive. Coverage isn’t the problem. Thin coverage is.
Being everywhere doesn’t cost you. Being everywhere thin does.
The conversion gap that should shape expansion decisions
The clearest number in the whole study: When a brand is cited in a category closely related to its core expertise, that citation turns into an actual named recommendation 46% of the time. In an unrelated, distant category, the conversion rate drops to 18% — less than half.
Existing authority functions like a bridge: it reaches the neighboring category. It doesn’t reach across town.
Existing authority is a bridge. It reaches next door — not across town.
The question to ask before you expand
Every expansion decision — a new topic, a new service line, a new market — comes down to one question: Are you trying to be found, or are you trying to be recommended?
Being a findable source is cheap and travels easily. Being the recommended answer requires proximity to what you already do well, and it requires depth once you get there — not a single appearance across five different questions.
Source: Kevin Indig, Growth Memo, analysis of SEMrush AI Visibility Toolkit data across 1,094 categories, January–June 2026 (US ChatGPT)
COMPLETE THE FORM TO
BOOK A STRATEGY CALL
"*" indicates required fields
COMPLETE THE FORM TO
BOOK A STRATEGY CALL
"*" indicates required fields
Subscribe and Share – WE APPRECIATE YOUR SUPPORT
