Every time “getting cited by AI” comes up in a content meeting, somebody says the same thing.
“Let’s just publish more blog posts.”
I’ve watched this reflex play out a lot. The writing itself quietly becomes the goal.
But the posts pile up, and ChatGPT and Gemini keep citing everyone except you.
So what separates the content that gets cited from the content that doesn’t?
There’s now a survey that answers this by content type — and the answer is usefully blunt. Let’s look at it.
The Citability Ranking, According To 500 Marketers
The survey comes from NP Digital, the digital marketing agency run by Neil Patel (the survey).
In May 2026 they asked 500 marketers which content types get cited or referenced most in AI search.
The top three:
- Original research: 82%
- Comparison content: 76%
- Rankings and best-of lists: 57%
just means data you produced yourself: you ran the survey, you did the analysis, and the number didn’t exist before you made it.
That 82% works out to four marketers in five putting it at the top of the citability list.
The Bottom Of The List Is Where It Gets Interesting
Now the other end of the table:
- Opinion and thought leadership: 16% (ninth place)
- Video content: 2% (dead last)
is the “here’s where the industry is going” genre — essays built on your own point of view.
So the heartfelt opinion piece scores 16%, and video barely registers at 2%. Rough.
But I think the axis separating top from bottom is actually simple: whether the AI could have produced it by itself.
Opinions and general explainers are things a language model can generate endlessly, for free. It has no reason to cite yours.
Trying to win citations with pure opinion is like haggling with a vending machine. The machine isn’t being rude; it’s just not built to negotiate.
Why Your Own Numbers Force The Citation
Flip it around: no model can conjure the results of a survey you ran on 500 real people.
When an AI needs information that isn’t inside it, it has to pull from an outside source. And a specific number pulled from outside tends to arrive with its origin attached.
Which makes original data something close to “information the AI has no choice but to cite.”
This lines up neatly with a study we covered on this blog before — the 100,000-prompt analysis finding that factual data, not search rank, decided which brands the AI recommended. Both arrows point the same way: own a primary source that lives outside the model.
Two Caveats Before You Trust The 82%
Two things to keep in mind, though.
First, this is a survey of what 500 marketers perceive, not a measurement of citation logs. “Feels citable” and “actually got cited” can drift apart. (And NP Digital is an agency that sells marketing — worth keeping in the back of your mind.)
Second, content type isn’t the whole game. How the thing is built — clear headings, explicit sources — is generally held to matter for citability too.
Still, “the less an AI can make it, the more it has to cite it” seems like an honest reading of how citation works. Maybe I’m wrong about how far the 82% generalizes, but the direction of the axis is hard to argue with.
Conclusion: Move Some Budget From Volume To First-Party Data
So the takeaway isn’t “blogging is dead.” It’s that part of the budget going into volume should go into numbers only you can produce.
And the scale doesn’t need to be NP Digital’s 500 respondents. Customer data, usage stats, a small survey — anything the outside world can’t get elsewhere counts as original research.
Two moves to start with: inventory the unique numbers your company already has (customer surveys, usage data, support-ticket patterns), then pick one and publish it as a piece others can reference — a small study, a comparison.
Once it’s out, measure whether the AIs actually start quoting your number back. That closes the loop.
The answer to “what should we make” was probably sitting in your own drawer all along. Worth a slot on the next content-meeting agenda, I think.
Sources
- NP Digital (Neil Patel), “Content Types That Perform Best in AI Search”, May 2026, neilpatel.com (in-house survey of 500 marketers on which content types are perceived as most cited/referenced in AI search: original research 82%, comparison content 76%, rankings/best-of lists 57%, opinion/thought leadership 16% (ninth place), video 2% (last place). Caveats: perception-based survey, not measured citation logs; the publisher is a digital marketing agency.)
- ppc.land, “Original research tops AI search citations at 82%, NP Digital survey finds”, ppc.land
