You've heard the advice. You might have given it yourself in a client kickoff last quarter: pick one topic, own it completely, and let the AI engines learn to associate your brand with that one thing and nothing else. Niche down or disappear into the noise. It sounds disciplined. It sounds like the kind of hard truth a consultant says while the room nods along.
I've sat in that meeting. I've probably run that meeting. For the better part of eighteen months, "niche down for AI" has calcified into gospel across generative engine optimization circles — the confident, repeatable advice that if you want real topical authority in AI answers, you shrink your footprint. Cover less. Go deep on one thing and quietly walk away from everything adjacent to it.
Then Kevin Indig ran the actual numbers. Not a hot take, not a LinkedIn post riffing off a hunch about what ChatGPT probably rewards — a real analysis. 1,094 categories. 283,215 citation observations, pulled from ChatGPT between January and June of this year. And what the data says does not match what the niche-down crowd has been telling you since last spring.
The niche-down advice that's killing your topical authority
Here's roughly where the advice came from. Somebody noticed that a handful of brands with tight, obsessive focus on one subject kept showing up in AI answers for that subject. The observation was real. The conclusion drawn from it wasn't. "This brand is narrow and it wins" quietly turned into "narrowness is why it wins," and that leap has been repeated in enough decks and newsletters over the last year and a half that almost nobody stops to check it anymore.
It's an appealing story because it's simple, and because it gives you permission to say no to things. Cut the topic list. Kill the blog categories that don't fit the "pillar." Tell a client their content marketing plan is too scattered and needs a haircut. Some of that can be good advice for other reasons — focus is real, thin content is real, indiscriminate content pruning is sometimes exactly right — but it's been sold under the wrong banner. The banner says breadth is the enemy of AI search visibility. The data says something closer to the opposite.
What Kevin Indig's data actually found
Indig's Growth Memo piece, published August 3, set out to answer a specific question: does topical focus make a brand more visible in AI answers? The dataset behind it is large enough to take seriously — over a thousand categories, close to 300,000 individual citation observations, all pulled from ChatGPT across a six-month window.
The headline finding is the one that should worry anyone who's been telling clients to prune their topic list down to a single lane: the penalty for spreading across topics isn't really a penalty for spreading across topics. It's a penalty for spreading thin. A brand that covers twenty subtopics with genuine depth in each one is not worse off than a brand that covers three. A brand that covers twenty subtopics with a paragraph on each is worse off — but that's a depth problem wearing a breadth costume.
Say that twice, because it's the whole argument. Shallow presence gets punished. Topical breadth does not. Those are two different variables, and the "niche down" advice has been treating them as one this entire time. For an industry that talks constantly about not confusing correlation with causation, generative engine optimization discourse missed this one for a year and a half.
None of this makes the original observation wrong, by the way. The narrow brands that got cited a lot really were getting cited a lot. It's the story attached to the observation that was wrong. Those brands weren't winning because they said no to other topics. They were winning because saying no to other topics is an easy way, maybe the easiest way, to accidentally end up with real depth on the one thing you kept. Depth was always the actual cause. Narrowness was just the shortcut some brands happened to take to get there, and shortcuts get mistaken for requirements more often than any of us like to admit.
Depth vs. breadth, industry by industry
The part of Indig's analysis that should actually change how you plan is the industry breakdown, because it wrecks the idea that there's one universal rule here at all. "Niche down" was always sold as a blanket instruction. The category-level data says the correct move depends entirely on which category you're in.
| INDUSTRY | WHAT THE BREADTH DATA SHOWS | WHAT ACTUALLY PREDICTS CITATION |
|---|---|---|
| Finance & real estate | Topical breadth is directly rewarded — brands covering more ground get cited more, not despite the spread but because of it. | Coverage across more subtopics, provided each one clears a real depth bar. |
| Legal & healthcare | Breadth barely moves the needle. The mention-breadth association for legal stays negative at -0.058 even at full topic coverage. | Repeat, deep mentions of the same subject — coverage alone doesn't substitute for it. |
| Most other categories (the 1,094-category average) | Shallow presence gets penalized regardless of how many topics a brand claims to cover. | Depth per topic, not topic count. Brands covering many subjects with real depth aren't punished for the spread. |
Finance and real estate are the categories where breadth is flat-out rewarded — covering more ground helps a brand's citation performance, not despite the spread but because of it. If you run content for a lender, a proptech platform, or an insurer, the "stay narrow" advice was actively working against you, and probably still is wherever it's still being followed.
Legal and healthcare run the opposite way, and the number here is the one worth sitting with: -0.058. That's the mention-breadth association for legal categories, and it stays negative even when a brand has full topic coverage. Read that carefully. It doesn't say legal brands should niche down — it says covering more legal subtopics, on its own, does nothing for citation performance in that category. What matters is repeat mentions, the same subject coming up again and again in places AI engines already trust. A firm can cover every practice area it wants. Coverage alone was never going to be the lever. Depth and repetition were the whole game the entire time, and no amount of adjacent topics substitutes for either one.
And then there's everything in between — the bulk of the 1,094 categories that aren't finance, real estate, legal, or healthcare. That's where the general pattern holds cleanest: shallow presence gets penalized, and topical breadth, on its own, does not. If your category isn't one of the four named above, the safest read of this data is the plain one. Stop worrying about how many subjects you cover. Start worrying about whether each one you've claimed would survive a stranger clicking through and actually finding something worth quoting.
The real definition of topical authority
So what is topical authority, if it isn't the size of the fence around your subject matter? It's the depth of what's inside it. That's a small distinction, and it changes almost everything about how a content plan should actually get built. It's also worth saying plainly: an AI engine that cites your brand once this quarter and never again didn't just witness a fluke of ranking. It witnessed a depth failure, and depth failures don't get fixed by adding more topics next to the one that failed.
“A prototype is not a product. A topic you mentioned once is not a topic you own.”
What to do next
Stop asking how many topics you should cover. That's been the wrong question for eighteen months. Ask instead how many of the topics you already claim to cover would survive an honest depth audit — the kind where you count how many pages on that subject would actually earn a repeat citation, not just a first mention. That single reframe does more for your AI search visibility than any amount of topic cutting.
You're not behind because your content plan covers more ground than a competitor's. You're behind if half of what you cover is thin enough that an AI engine has nothing left to actually quote. Those are different problems, and only one of them gets fixed by shrinking your site. Go wide if your category rewards it. Go deep everywhere you go. That was always the actual advice — it just got translated badly somewhere on the way to the deck.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.