How to Create Buyer Personas for AI Search

TL;DR: Buyer personas for AI search need to be useful beyond a slide deck. Start with the usual buyer research, then map the questions each persona is likely to ask throughout the buying journey.

I have a confession: I used to think buyer personas needed names. You know the ones? Marketing Mary. Founder Fred. Scaling Sam. 

You can’t see me, but I’m cringing. Hard. While I still think buyer personas are useful, I don’t really care whether we know that Sam is 42, listens to podcasts on his morning commute, and enjoys hiking on the weekends. What I care about is what Sam would ask ChatGPT when he’s trying to solve a problem.

That’s the shift AI search has created for buyer personas.

The traditional audience research still matters. You need to understand who your buyers are, what they’re trying to accomplish, what frustrates them, what might stop them from buying, and who else is involved in the decision. But if you’re building an AI search visibility strategy, your personas need to go one step further.

You need to understand what each buyer would actually ask, how those questions change throughout the buying journey, and how the context surrounding that buyer could affect the answers they receive.

In other words, your buyer personas need to become operational.

Key takeaways

• Think of your buyer personas as prompts. Map what each buyer would realistically ask an AI system instead of stopping at demographics and pain points.
• Account for the buying journey. The questions someone asks while identifying a problem will be different from the questions they ask while comparing vendors.
• Cover question categories, not every possible prompt. Focus on commercially meaningful questions tied to personas, pain points, buying stages, and use cases.
• Measure visibility by buyer, not just overall. Personalization can affect which brands AI systems surface, so one generic prompt set may not tell the whole story.
• Use visibility gaps to guide your content strategy. If your brand disappears for an important persona or buying stage, investigate what information, evidence, or authority is missing.

What Is an AI Buyer Persona?

An AI buyer persona is an extension of the traditional buyer persona that accounts for the questions and context a buyer might bring to an AI system while researching a problem, comparing solutions, or making a purchase decision.

A traditional persona might tell you:

• The buyer’s role
• Their goals
• Their pain points
• Their objections
• Their buying triggers
• Who else influences the purchase

Keep all of that, but for AI search visibility specifically, ask: What would this person actually ask an AI assistant?

Let’s say you’re selling CRM software to B2B SaaS companies. Knowing your buyer is a VP of Sales at a 20-person company is helpful. But that buyer isn’t necessarily going to open ChatGPT and type a generic prompt like, “best CRM software.” They might ask: What’s a CRM for a 20-person B2B SaaS sales team that integrates with our existing tools and doesn’t require a full-time admin?

Now we have something much more useful. We have the persona, situation, constraints, and commercial intent wrapped into the question. And that gives us something we can actually use to measure AI visibility.

Why Traditional Buyer Personas Aren’t Enough for AI Search

I started thinking differently about personas after a conversation with Lindsay from Conductor on the Found in AI podcast. After we recorded the episode, I gave listeners an experiment: “Stop thinking about your personas as slide deck profiles. Instead, start thinking about them as prompts.”

I used runners as an example. Imagine you sell running shoes. One of your customers is a beginner who signed up for a Thanksgiving 5K. Another is an experienced marathon runner. They’re both runners, and they could both be looking for running shoes. But are they going to ask ChatGPT the same questions?

Probably not.

The beginner might be worried about comfort, injury prevention, price, or whether they even need a specific type of shoe for their first 5K. At the same time, the marathon runner could care about mileage, weight, durability, race-day performance, or how a shoe compares with one they’ve already used.

It’s essentially the same product category, just with very different information needs. And that’s what gets lost when a persona lives on a slide somewhere and never makes its way into your AI search strategy. Demographics tell you who someone is, but they don’t necessarily tell you what that person is going to ask.

AI Search Personalization Changes the Visibility Problem

There’s another reason I’m pushing brands to think more seriously about the people behind their prompts. AI answers aren’t necessarily the same for every user.

In 2026, I worked with João da Silva at friction AI to study how user context affects brand recommendations across ChatGPT, Gemini, Claude, and Perplexity. We tested six controlled personas across three consumer categories and 30 prompts, resulting in 8,609 successful collections over nine days.

We compared three conditions: 

• Cold web-enabled API sessions with no user history
• Temporary logged-in sessions without retained history
• History-primed accounts containing controlled user information

We saw the biggest difference with Gemini. The brands recommended to our history-primed users changed 33.3 percentage points more than we would expect from the normal variation between AI answers. For ChatGPT, that difference was 16 points. Perplexity showed a smaller 7.2-point difference. We didn’t detect the same recurring change in Claude’s brand recommendations, although we did see its search behavior change.

In other words, two people can ask an AI system about the same product category and encounter different sets of brands based on the context the system has about them. Personalization makes a big difference.

That’s important because it changes the question marketers need to ask. Instead of: Does my brand show up in ChatGPT? We need to start asking: Does my brand show up for the buyers I actually need to reach?

Geography gives us a good example. The personas in our study were based in the UK. When the AI systems had access to that context, UK and European brands appeared more often, U.S. brands appeared less often, and the systems cited more .uk sources. The exact numbers were a 6.6-point increase for UK and European brands, a 4.9-point decrease for U.S. brands, and a 6-point increase in .uk citations.

So when I talk about buyer context, I’m not talking about creating a 12-page profile detailing someone’s hobbies and favorite coffee order. I’m talking about information that could actually affect what they’re looking for and which brands make sense for them.

This doesn’t mean you need to create a separate persona for every possible person who could buy from you. Please don’t do that. It means one generic prompt isn’t necessarily representative of what all of your customers will encounter in AI search. And that’s why buyer context needs to be part of how you design and measure an AI search visibility strategy.

Start With the People Involved in the Buying Decision

If you already have buyer personas, you don’t need to throw them out and start over. Start with the research you already have.

• Who buys your product? 
• Who uses it? 
• Who influences the decision? 
• What does each person care about?

For a consumer brand, those differences might come from experience level, budget, location, intended use, preferences, or the problem someone is trying to solve. B2B can get more complicated because one purchase may involve an entire buying committee.

Shane Tepper, a guest on Found in AI, explained this really well when I asked him why his team evaluates AI visibility across such a large set of queries. Essentially, a CFO and an IT leader can be evaluating the exact same product while asking completely different questions. The CFO may care about ROI. The IT leader may want to know whether the product integrates with the company’s existing tech stack. 

As Shane told me, “You have to have queries for all of these personas.”

That’s the part I want marketers to pay attention to. One company can have several buyers researching the same purchase from entirely different perspectives. Your persona work should capture those differences before you start building a prompt library.

Turn Each Buyer Persona Into Commercial Questions

Once you know who is involved in the buying decision, map what each person could ask throughout that decision. I like to think about it roughly as:

1. Problem identification
2. Education
3. Solution exploration
4. Comparison
5. Validation
6. Decision/action

The terminology doesn’t have to be precious here. Your customers probably aren’t thinking, Ah yes, I have officially entered the validation stage of my journey. You’re simply trying to understand how their questions change as they get closer to a decision.

Someone identifying a problem might ask: Why is our sales team spending so much time manually updating customer records? As they learn about possible solutions to their problems, their questions might evolve like this:

• What types of CRM automation could reduce manual data entry for a small sales team?
• Which CRMs have strong automation features for B2B SaaS teams?
• How does Product A compare with Product B for a 20-person SaaS company using [existing technology]?

The questions should change as the buyer moves closer to a decision. Someone who is just figuring out why their sales process is such a mess needs different information than someone who’s already narrowed their CRM options down to two vendors.

This is where the work you’ve already done on your buyer personas becomes useful. Look at their pain points, objections, buying triggers, decision criteria, and product requirements.

• What would they actually want to know about each one? 
• What would make them start looking for a solution? 
• What would they need to know before putting a product on their shortlist?
• What could stop them from buying once they’re close to a decision?

You don’t need to predict every possible way someone could phrase those questions, either. We’re trying to understand the types of questions that come up during the buying process, not brainstorm 200 variations of “best CRM software.”

Shane described his team’s approach as a matrix built around persona, buying stage, pain point, and product features or use cases. The goal, he explained, is to cover the potential questions—or at least the question categories—a buyer might ask while comparing solutions. 

That’s a much better way to think about prompt research. You’re trying to represent the commercial questions your buyers could ask, not predict every sentence they’ll ever type into ChatGPT.

Build a Persona × Question Matrix

Once you’ve done that work, put it somewhere your team can actually use. It does not need to be fancy. For a B2B software company, a simplified version could look something like this:

From there, you can expand the matrix with the stages, concerns, use cases, and question categories that make sense for your actual customers.

Use Your Personas to Measure AI Search Visibility

Now the persona becomes part of your measurement strategy. Take a representative set of commercially meaningful questions from your persona matrix and test them across the AI systems your customers are likely to use. Then look at what happens. For each persona and question category, I want to know things like:

• Does the brand appear?
• Which competitors appear?
• How is the brand described?
• Which sources are supporting the answer?
• Does the recommendation change at different stages of the buying journey?
• Are there personas for whom the brand is consistently absent?
• Are there important capabilities the AI system misunderstands or overlooks?

This is similar to the approach Shane described on Found in AI. His team runs its query sets through ChatGPT, Perplexity, Gemini, and Claude, then analyzes how clients perform against competitors across different personas, buying stages, pain points, features, and citation sources. 

That’s much more useful than throwing 50 generic prompts into a tracking tool and averaging the results together. If you lump together all possible prompts a buyer might ask an AI engine, you can miss the part of the buying journey where you actually have a visibility problem.

Buyer Personas Should Shape Your Content Strategy, Too

Your persona research should eventually tell you what to create, update, or clarify. For example, if you’re visible for practitioner questions but absent when CFOs ask about ROI, look at the information supporting those questions. Do you have credible ROI data, customer results, or third-party validation? If you appear during early research but disappear when buyers start comparing vendors, review the content and sources available to help them make that comparison.

Use the gaps you find to prioritize your content strategy. You might need a new asset, stronger evidence for an existing claim, clearer product information, or better third-party authority. The AI persona simply helps you figure out what information your buyers need and where your brand isn’t providing it yet.

Don’t Create One Generic “AI Persona”

You don’t need to create a separate “AI persona.” Start with the real people you’re trying to reach. For each buyer, consider:

• What they’re trying to accomplish
• What could affect their decision
• What they would actually ask as they research their options 

Our Personalization Gap research shows why that context matters: AI systems can surface different brands based on what they know about the user. Use those buyer-specific questions to test what AI systems are actually telling your customers and where your brand fits into those answers.

Your Buyer Personas Should Tell You What to Measure

I still believe in buyer personas. I just don’t think a fictional name and a list of demographics are enough anymore. For AI search, a useful buyer persona connects customer research to the questions people ask, the context surrounding those questions, and the information they encounter while making a decision.

It can shape your prompt library and can make your AI visibility measurement more representative of your actual customers. And, most importantly, it can show you where your content and authority aren’t supporting the questions that matter to the buying decision.

That’s also how I approach AI Search Visibility Audits.

I don’t want to know whether your company appears for a random pile of prompts. I want to know whether you’re visible when the people who could actually buy from you are researching their problems, evaluating their options, and deciding what to do next. Because “we show up in ChatGPT” isn’t a particularly useful business metric if you’re invisible to the buyers who matter.