- Jun 11
- 7 min read
Every few months, a new acronym drops into the SEO world and people scramble. AEO. GEO. LLMO. Take your pick. Our industry has a talent for making old things sound new, and the AI-search era has been particularly generous here.
Here's what I tell our clients and what I keep seeing prove true: brands consistently winning in every algo update or AI search development are the brands that know, deeply and specifically, who they're talking to. That’s your ideal customer profile, or ICP. Yes, it’s another acronym. But in this article, you'll see why I think it’s the most important one.
What’s an ideal customer profile (ICP)?
Your ICP is the specific type of company or person most likely to buy from you, stay with you, and get real value from what you do. Your ICP is not a vague persona, but a precise profile built from your actual best customers. And it's a crucial part of SEO for B2B marketers.
Most teams I talk to have an ICP document. And most of those documents are sitting on a server somewhere collecting dust. In my experience, its use is often limited to marketing briefs and new hire orientation sessions—created, filed, and promptly ignored.
Meanwhile, teams build content strategies based on what the CEO thinks the customer wants, or on what competitors are ranking for. But a data-backed ICP tells you exactly what your actual customers need, and it's a foundational piece of your marketing playbook that holds up across both classic and AI search.
Your ICP isn't a slide in your sales deck. It's the north star that makes every content decision clearer, every experiment smarter, and every algorithm shift a lot less panic-inducing. And if you've been in digital marketing for any length of time, there's a decent chance you've quietly let it go stale.
Why it's important to know your ICP for AI search
AI search platforms receive 8 billion visits a month. These systems evaluate whether a piece of content genuinely satisfies the intent behind a query. And they do it for each user individually, taking factors like chat history into account. AI systems are doing, at massive scale, what a great human editor would do: distinguish content written to help from content written to rank.
Yext's analysis of 6.8 million citations across Gemini, ChatGPT, and Perplexity found one thread running through all three: content that knows its audience, speaks directly to them in the right context, and demonstrates real-world experience tends to get cited more often. See also: 10 ways customer loyalty drives visibility in personalized search.
Meanwhile, content written to appeal to everyone tends to have to have a more difficult time due to how generic it is. Gemini, ChatGPT, and Perplexity each source content differently when it comes to structure or nice domain authority, but none of that variation matters if the content itself isn’t built for a specific purpose.
There's a meaningful upside to being cited, too. At my digital marketing agency, 97th Floor, we see AI-search traffic convert at a much higher rate. The opportunity or volume is often much smaller but dramatically more valuable and the competition often melts away the more specific you get. Plus, 93.2% of marketers report that personalized/segmented experiences have led to more leads and purchases, according to HubSpot.
This is where Google's framework of Experience, Expertise, Authoritativeness, and Trustworthiness (EEAT) went from abstract principle to concrete competitive advantage. The extra "E" for Experience is what I recommend marketers should focus on. AI answer engines aren't just looking for credentials in a bio. They're reading for specificity and first-hand experience in the prose itself. Have you done this? Have you seen this? Is this written for a real person with a real problem?
While search marketers have been aware of EEAT for some time now, the last decade of SEO was mostly a triumph of process over people. Keyword research drove editorial calendars. Content teams were handed spreadsheets of search volumes and keyword quotas and told to write. The question shifted from "what does my customer actually need?" to "what does the algorithm want?"
AI is forcing that to change. Generic content can no longer hide behind backlink velocity or word count. The content volume game created an enormous amount of noise, and AI systems are the best noise-canceling headphones ever built.
How to identify your ICP
At 97th Floor, the best ICP research we've ever done wasn't in Ahrefs or SEMrush. It was through listening to recorded sales calls or in 30-minute conversations with a client's best customer, who told us about their experience before finding a solution, what language they used to describe their problem, and insider info that someone who hasn’t been through the situation wouldn’t know. That single conversation could be worth more than weeks of analytics data.
In marketing, it can be super easy and even tempting to just sit behind the laptop. But when did you last talk to your customers? Not survey them. Not segment them in your CRM. Actually talk to them.
Here's how to rebuild that foundation.
Analyze your real customer data
Your proprietary customer data can give you insights that help you convert customers from AI and other marketing channels. Marketing personas based on 3rd party data have their place but your primary data is incredibly valuable.
Take a look at your best real customers, not hypothetical ones.
You can of course consult data for AI search conversions within Google Analytics but the real gold is in the conversations behind the data; support tickets, sales call notes, customer interviews, and comments on social posts.
Pull marketing data on your highest-LTV, lowest-churn accounts and ask what they have in common; not just by industry or company size, but by how they think about their problems and how they make decisions.
Conduct real interviews
When you interview your customers, you can identify what really matters for them. This can shape which intent you target in your AI content strategy.
Five to ten interviews is enough to start. Cover six areas:
What triggered their search
How they researched options
Who was involved in the decision
What they found valuable
What their real challenges were
Where they go for industry information
Many of their answers will surprise you. They always do.
Mine your existing content for signals
Which web pages have the highest engagement, lowest bounce rate, and strongest conversion? Those aren't accidents. They're showing you what your ICP actually responds to when it shows up. Use this to build your briefs, map out your content calendar, or decide what content to atomize across social channels.
Even better, take the value propositions, CTAs, and subjects that actually engage your audience, and share them across the rest of the org. This is exactly the kind of information that typically stays siloed within an org but could have tremendous impact.
Copy and paste your customers' language
The exact phrasing people use to describe their own problems (before they've been educated on your solution's terminology) is the phrasing that resonates in content. Don't rewrite it into marketing speak. Use it verbatim.
This kind of writing will mean you’ll more closely match the queries users are searching in AI, it will better demonstrate to your audience that you speak their language (it’s literally their language), and it’s a great way to cut through the noise especially in the B2B space.
How ICP knowledge changes your content strategy
Absolutely. Building depth around a topic through multiple pieces of interconnected content still signals quality and earns trust signals across platforms. Content clusters, internal linking, and consistent topic coverage still work. None of that is dead.
What changes is the anchor. A content cluster built around "email marketing" for a generic audience is getting outcompeted by content built around "email marketing for mid-market SaaS companies managing long sales cycles." That specificity isn't just for humans, it's what helps AI search understand who your content is actually serving. That context is how you get cited instead of skipped.
Here's what else shifts when you put ICP first:
Your content brief starts with a persona, not a keyword. The keyword and topic still matter, it's just the second decision, not the first.
Who is this for?
What do they already know?
What do they need to believe to act?
Those questions come before you open a keyword tool. Once you finally reach for some keyword research, you’ll find that many of the ideal queries will have zero search volume.
That’s expected and is increasingly the point. Keyword tools in this strategy shift from traffic predictors to validation and discovery tools. They can help to confirm that a topic exists, adjacent terms and ideas, and to identify patterns across site or groups of pages. It’s always been true, but showing up for a topic or keyword with 20 monthly search on average from the exact buyer you’re trying to reach is worth more than an extremely broad term unrelated to your product or service that has a volume of 20,000 per month.
In this world, long-tail queries have become more valuable, not less. AI handles the broad stuff. The opportunity for specific, high-intent content has grown because AI seems to favor answers made to precise questions, and the brands answering those questions best are the ones being cited and mentioned most often.
The algorithm will change. Your customer won't.
Every few years it seems that something upends search. Panda. “Not provided” keyword data in Analytics. Mobilegeddon. And now, AI. The teams that get rattled are the ones that built their strategies on tactics alone. The teams that adapt fast are the ones who built on customer truth, because that doesn't change when the platform does.
Your ICP is the one constant while everything else shifts. It makes every strategy decision cleaner, every experiment more useful, and every new acronym less threatening. You don't need to predict what AI search will do next if you know your customers well enough. So, when's the last time you talked to them?
