Search marketing has always been a venue where small companies could punch above their weight. At Recurrent, where I’m head of marketing, we outperform companies 500x our size with a simple formula: Proprietary data → compelling research articles → media coverage → SEO performance → more proprietary data.
But that’s SEO. AEO has been more of a slog.
Not only are we in an emerging category without standardized language, but we also support a variety of stakeholders in the auto industry. That means we are different things to different people. While SEO allows for that adaptability, AI models favor consistency that is difficult for us to offer: consumers on Reddit say one thing, industry publications say another, all while we try to shape the new category language with our website.
This newsletter is another example. Almost no citations despite some decent SEO. First to Market isn’t covering one TYPE of business. It’s a LENS on business: we’re all building in new categories that aren’t fully defined, where the tried-and-true playbooks for other companies don’t really work. We all get that as startup people, but it’s a tricky concept for an LLM.
I came up through SEO, and consider myself more adept than most at AEO, and it’s still difficult for me. Two experts and fellow startup leaders offered to help.
Esteve Castells is cofounder of LLM Pulse.
Luca Fancello is cofounder of Qwairy.
Before we get into ‘how,’ let’s start with ‘why’ AEO requires a different playbook for us.
What’s different in a new category
Being small is an advantage. No brand has built up an insurmountable lead in a new category so a few citations can have an immediate impact, as one study found.
You’re not filling a gap. In a mature category, you are looking for the places that incumbents missed or are not positioned to answer. A new category is more like creating boundaries because a lot of times the responses being returned today are sourced from incumbents in adjacent categories. That’s another example of the findability problem.
Volume is different than demand. Search platforms are designed for existing volume in defined categories. But we’re often starting from near zero here. “Keyword tools will tell you the market does not exist. They are measuring a name that has not been adopted yet, not demand,” said Esteve.
The language is in flux. One study found that Gymshark beat bigger incumbents (with better SEO authority) on athleisure topics by nimbly mirroring the phrases that consumers put in their AI prompt.
What doesn’t change
Buyers still describe problems. Whether someone is looking for a more affordable CRM or a formal shoe that fits wide feet, they are describing a problem. If you understand the customer, and can anticipate their language, then you can create content around the solution.
Traffic from AI chat still converts. LLM referral traffic converts as well as any other source. Lots of people have covered this, and I can validate it from personal experience. In a new category, where education is a critical part of the sales process, that could be even more impactful to us.
Third-party mentions still matter most. An LLM builds context by synthesizing across the web: every brand mention, Reddit comment, news story, product review. Your own site is just one of many, many, many signals.
Start here with AEO
Whether you’re deciding to revive an existing category or create a new category will ultimately influence your long-term AEO path. But that won’t matter until you’ve created a solid foundation. Either way, we all start with these steps.
1. Start from the buyer’s problem.
Our category name probably isn’t in anyone’s prompt yet. The prompts that people are using more likely describe the pain or ask for an alternative to their current attempted solution.
Esteve suggests: “Write down the 10 to 15 prompts a buyer would type. Those are your market until the category name exists.”
Test it: Try to draft the customer prompts that would have preceded your last 10 sales. Be SUPER specific with them then look for overlap at the end. Note: this exercise is even more revealing if you have call transcripts or long email threads.
2. Give the model a consistent summary.
“Update your one-sentence [company] description and repeat it identically everywhere,” said Esteve. “The model is going to paraphrase… [so] give it one sentence to paraphrase instead of five variants.”
This doesn’t need to replace your company slogan or homepage hero. It’s just the boilerplate summary of what you do for people that gets included in LinkedIn pages, media coverage, around the website, etc.
Test it: Go back to the prompts you wrote in Step 1 and use the common language + some aspirational positioning. That’s your first draft.
Common misstep: Don’t be cute. We all took a creative writing course and this isn’t that. It’s a mirror of the language your customers use.
3. Find the future mentions.
“In an emerging category the model has almost nothing to cite except your website, and it discounts your website because it reads as marketing,” Esteve told me.
Luca added: “A third-party mention beats self-praise every time [because] an AI discounts you calling yourself the best.”
Getting media coverage in a category that nobody covers is not fun. There’s no beat reporter or niche trade publication so it’s mostly pitching to people who don’t know or care why what you’re doing matters. (I’ve got hundreds of unanswered pitches in my Sent folder.)
But the other side is that nobody else has won it either. The bar is low and that makes it more likely for a no-name startup to clear it.
“Three or four placements can visibly change your share of voice in weeks,” Esteve confirmed.
Test it: Get an idea of the places that you need to show up by running the 10 customer prompts (from Step 1) and writing down the sources that get cited. You don’t need THOSE placements, just placements like those.
4. Give them something to cover.
Pitching a product to journalists is the worst PR sin. I do a lot of PR and meet with a lot of journalists and they often tell me about how they get (literally) hundreds of terrible cold email pitches per day.
Pitching your product is about you. But pitching exclusive data or information that they can’t find anywhere else is about them because you’re doing the first half of their job for them.
“Original data, a ‘state of the category’ study, the glossary that names the space,” Luca said. “Data and studies get cited constantly and can’t be copied.”
Test it: What can you see that nobody outside your company can? Aggregated data, consumption or usage patterns, customer surveys, etc. Those are the things that create a story that only you (and your company) can help to write.
Common misstep: This is not the time to look for a short cut. “New brands hunt for the trick that games the LLMs this week… when the thing that actually gets you cited is boring and slow,” Luca said.
5. Track your progress.
The main takeaway from Esteve: “When someone describes the problem you solve, who does the model name?”
That’s ultimately what we’re optimizing toward. You don’t need to spend $1,200/year on a platform starting today — there’s probably nothing to track yet. But once you’ve worked through Steps 1-4, you should have systems in place to track your ROI. Both contributors here build tools for exactly that (LLM Pulse and Qwairy) and there are others.
What comes next
Your AEO playbook after Step 5 depends on whether you’re attempting to win inside an existing category or attempting to build your own.
Category creation is a major positioning decision. AI search doesn’t really change that calculus.
But to help startup leaders going down both paths, I’m tracking companies in the category database against a fixed set of prompts. We’ll have a ton of great data in a month so expect an update once I have it.
In the meantime, a huge thanks to Esteve and Luca for contributing. They are super smart so we’re lucky to tap into their wisdom.
From the Workbench this Week
I’m running Steps 1-5 on this newsletter right now. As you can see, there’s some work to do.
The Category Signals page is half-baked right now. It’s a unique dataset and something that other newsletters don’t have or don’t give away for free. My goal this week is to continue to expand it and make it more helpful to both existing subscribers and future media mentions.
Ask me again in a quarter how it’s going.
Have a beautiful week.
Loved learning the nuances of how AEO applies to another use case versus the spray and pray broad advice. "Don't be cute" is perfect framing when describing your company. It's a fun challenge, at least for me, to say something different enough but recognizable too.
With a new category, AEO feels like something you have to keep working on instead of just making a strategy and leaving it. You need to keep checking what customers are asking, what is getting picked up, and what needs to change based on that. And when the language around a category keeps changing, how often do you think a new category should go back and check these signals?