AI search. The term sparking 3 acronyms, 6 LinkedIn guru frameworks, and at least 1 new online course (yours for only $999!) every other day.
Let’s cut through the noise here.
We sat down with Noah Greenberg, CEO and co-founder of Stacker, a content syndication platform, to find out what actually earns a brand a spot in AI answers.
The TL;DR is that AI search reconstructs trust from signals scattered across the web — who's getting mentioned, cited, and repeated. And visibility really comes down to one question: "Does the internet, in aggregate, treat my brand as an authority?"
This presents a major opportunity for brands who operate like media companies (ahem, something we talk about a lot here at the ‘arb).
Noah spotted this early through Stacker's customers, who started noticing that when their content was republished on third-party outlets, it appeared more frequently in LLM responses.
Keep reading for how Noah experiments with AEO, the quick 4-step test you can run yourself, and why the brands earning AI citations are the ones publishing what nobody else can.
Key takeaways:
- Content republished on third-party outlets appears most frequently in LLM responses.
- The best brands create information nobody else can, whether that’s proprietary data or genuine expertise.
- You don’t need a massive strategy to get started. Pick 1 customer and 1 problem, and run 1 small content experiment.
- AI models cite sources the same way readers do: They follow whatever the broader web treats as authoritative.
The best brands are starting to look like media companies
Breaking through is harder than it's ever been, and it’s not just the algorithm’s fault. Noah points to three major shifts:
- Paid channels are getting more expensive and less effective.
- Audiences are spread across more platforms.
- AI has made it easier than ever to flood the internet with generic content.
The brands winning attention are investing in people who know how to tell stories. Some buy media companies, like HubSpot’s acquisition of The Hustle. Others build a newsroom with journalists, or partner with subject matter experts.
Noah sees his own LinkedIn presence as the same strategy, just at a different price point:
I'm just doing the poor man's version of buying The Hustle for 30 million bucks.
He adds, “You need to become an authority in your customer's mind, so that when you want to sell them something, they trust you.”
It sounds obvious: People don’t want to be sold to. You build an audience by creating something people actually want (teach them something, entertain them, make them better at their job).
…and yet, the vast majority of B2B content is self-serving, ICP-blind, and a little dull. At best, it’s thinly disguised product marketing.
You earn the right to talk about your product by not talking about your product.
Turns out, that approach is extremely valuable for AI search, too: Every time you publish something useful and it travels, you're feeding the models another signal that your brand is the authority. Do it consistently and you become the name that comes up to your customers and the chatbots they're asking.
Building a brand like this is playing the long game. The good news is, you can start playing the long game now, with early tests you can run in a weekend. Here's the 4-step plan.
A 4-step plan to go from 0 to your first AI visibility test
AEO and GEO can feel intimidating, especially for marketers who don’t come from technical backgrounds. But Noah argues the highest-leverage work is much more accessible.
“Go spend a weekend, or 30 minutes every morning for 2 weeks, learning about GEO,” he says. “Most of GEO is something an English major can do.” (Editor’s note: No offense to any English majors out there, us included.)
The starting point is understanding how people ask questions, what information AI models surface, and what kind of content earns their trust.
Step 1: Study the basics
There’s no shortage of people opining on AI search. Start by finding 3 different GEO 101 explainers across formats: YouTube videos, newsletters, blog posts, podcasts, whatever works for you. Pull out the ideas they all agree on.
The goal isn’t to become an expert overnight. It’s to understand the foundational concepts before you start testing.
Noah’s go-to resources include:
Step 2: Experiment like your customer
The biggest shift in search behavior is queries moving beyond keywords and into real conversations.
Conversation“I’m a 30-person company outgrowing our spreadsheets but my team will revolt if it’s complicated. What’s a CRM they won’t hate?”
Search“project management software”
Conversation“We’re a design agency drowning in client feedback across email, Slack, and Figma. How do other agencies keep this organized?”
Search“email marketing tools”
Conversation“I send a newsletter to 5k people from Mailchimp but the deliverability is getting worse. Is it me or the platform?”
Search“payroll software for small business”
Conversation“We just hired our first 3 employees in different states and I have no idea how to handle payroll taxes. Help.”
Search“accounting software”
Conversation“I’m a freelancer who’s been doing my books in a notebook and tax season is destroying me. What’s the simplest thing that works?”
To start seeing what this shift means for you, ask ChatGPT the questions your customers would ask. Then ask Claude the same things. Then ask Gemini the same things.
You’re looking for patterns in the way AI interprets the problem and what information it surfaces.
Pick one customer persona and brainstorm 50 things they actually talk about in your space: frustrations, questions, half-formed problems, and challenges they might not know how to solve yet.
You can see how this plays out in full with the “best CRM” search:
While in there, listen for what kind of content your customer actually wants — because that tells you what specific content is actually worth making.
As you run your prompts, you'll notice your customer is usually looking for looking for 1 of 2 things:
- They want to be educated and entertained. They used to get this from trade publications and magazines. You win here with real expertise and a point of view.
- They want proof to back up their own decisions. They want numbers nobody else can publish. You win here with proprietary data turned into research, reports, and trends.
The brands most likely to earn an AI recommendation are the ones creating information nobody else has: original data, unique expertise, firsthand experience, and perspectives built from doing the work. A couple examples:
As Noah puts it: "These LLMs are looking for unique, authoritative content." That matters even more in an AI world, where anyone can generate something that sounds good in seconds.
Step 3: Run 1 small test
“Don't boil the ocean,” says Noah.
Your goal in the first 3 months is not ‘We’re going to double LLM traffic.’ Pick 1 bite-sized niche, 1 part of the funnel, and say, ‘We're going to try and impact only that.
His favorite piece of low-hanging fruit: Show up when someone describes a problem your product solves, not just when they ask for the best option in your category.
To carry on with the CRM example, that means targeting a specific problem like the spreadsheet-revolt conversation.
What that test looks like depends on the lane you picked in Step 2:
- If you're the "data" brand: Take 1 proprietary number you can publish and nobody else can (say, the average time your customers spend on a CRM before they outgrow it, or how adoption rates change by team size). Write it up as a small piece of original research tied to a single customer problem. Then watch whether models start citing your figure when someone describes that problem.
- If you're the "expertise / replacing magazines" brand: Take 1 half-formed problem (say, “we've outgrown spreadsheets but every CRM looks like overkill”) from your list of 50 and publish the genuinely useful, opinionated answer your customer used to get from a trade pub. Then watch whether models surface your take when someone asks about it conversationally.
Step 4: Measure what changes
Repeat after Noah: “You can’t improve what you aren’t tracking.”
But don’t overcomplicate tracking, either.
“Sign up for the cheapest version of 1 of the dozen AI analytics tools out there,” he says. “Don’t spend more than $200 or $250 a month on it. They all do basically the same thing.”
What you’re looking for is simple: where your brand appears, where your competitors appear, and what types of pages or URLs are being cited for the prompts you care about.
Those patterns will tell you what content to create, update, and test next.
P.S. Want to track how models actually talk about you over time? Our LLM sentiment tracker template makes it a copy-paste job.
⚒️ LLM Sentiment Tracker Template
Download our template for tracking LLM sentiment about your company, industry, and key topics over time.
AI visibility is a specificity game, not a volume game
There’s content only you could put out into the world. That’s the content that will get you seen today.
AI search reconstructs trust from whatever it can find about you across the internet. The way you feed it is the same way media companies have always earned an audience: say something useful, entertaining, and timely — and say it often enough that you become the name that comes to mind.
AI models are just one more reader now, learning who to trust from the same signals everyone else does.
At the 'arb, we've always believed good old-fashioned storytelling wins in the end. Noah's customer data — and a growing pile of earned-media research — now happens to agree.
So don't ask what the algorithm wants. Ask what only you could publish. Then go make sure it travels.
FAQ
What if my brand doesn’t have proprietary data or SME access?
If you don’t have original data, lean into expertise. If you don’t have a recognized SME, build one. That’s what Noah’s doing with his own LinkedIn presence. The bar isn’t “we have a research team." It’s “can we publish something useful that nobody else would think to publish?”
What if we’re a small team with limited content bandwidth?
This isn’t a volume game. Pick 1 niche, 1 part of the funnel, 1 test. 1 well-researched piece tied to a specific customer problem will outperform 10 generic posts. Limited bandwidth forces the specificity Noah’s strategy requires anyway.
How do I know if my test is actually working?
Run your target prompts into ChatGPT, Claude, and Gemini before you publish anything and document what comes up. Repeat every few weeks after you’ve published. An AI tracking tool can systematize this if you’re tracking multiple prompts, but per Noah’s recommendation don’t spend over $200-$250/month.
What kinds of content are most likely to travel on their own?
Named data points, original research, and concrete claims. These get cited because they’re specific and sourceable. Noah’s customers saw this directly: The content that showed up in LLM responses was the content that had already been picked up and republished. The models were following the same signals as human readers.
