How to do keyword research for GEO: appear in AI answers like ChatGPT and Gemini

Table of contents

1. What is keyword research for GEO?

2. Step 1: Start with your buyer persona’s real pain points

3. Step 2: Understand intent vs. literal prompt matching

4. Step 3: Group common questions by intent

5. Step 4: Run a visibility analysis in AI search engines

6. Step 5: Prioritize prompts by value, volume and opportunity

7. Step 6: Create content that LLMs want to quote

8. Conclusion: Keyword research for the age of AI search

What is keyword research for GEO?

Traditional SEO starts with keyword research:
What are people searching for on Google?

But the rise of AI search engines like ChatGPT, Gemini, Claude, and Perplexity is changing how people access information.

These tools don’t return ten blue links. They generate a single, confident answer.

So now the question is:

How do you show up in that answer?

That’s where Generative Engine Optimization (GEO) — also known as Large Language Model Optimization (LLMO) — comes in.

And it starts with a new kind of keyword research.

Step 1: Start with your buyer persona’s real pain points

Forget keyword volumes for a second.
Start with the real problems your ideal customer is trying to solve.

In AI search, people don’t type keywords — they ask questions.

Instead of: CRM for startups
They might ask:

“What’s the best CRM if I’m a solo founder with no tech team?”

This is the AI equivalent of a long-tail keyword — a natural language query with clear intent.

Step 2: Understand intent vs. literal prompt matching

Here’s the key difference between SEO and GEO:

LLMs don’t look for keyword matches. They interpret meaning.

So you don’t need the exact prompt to rank.
You need to understand the intent behind the query.

Example:

These three prompts are different but lead to the same kind of answer:

• “What’s the easiest CRM for small businesses?”

• “Best CRM for non-technical entrepreneurs”

• “Can you recommend a simple CRM for freelancers?”

Same intent, different wording.
This is why intent modeling matters more than exact phrasing in GEO.

Step 3: Group common questions by intent

To build your prompt research map, cluster questions around shared motivations or problems.

Let’s say you sell accounting software for freelancers. Here’s how that might look:

Pain 1: Just getting started

– “How do I manage taxes and invoices as a freelancer?”
– “Simple accounting tools for self-employed beginners”

Pain 2: Losing time using Excel

– “Alternatives to Excel for tracking expenses”
– “Why Excel isn’t ideal for freelance accounting”

Pain 3: Growing the business

– “Scalable invoicing software for freelancers with multiple clients”
– “What tools help freelancers manage high-volume billing?”

Each group represents a prompt cluster based on intent — not exact keywords.

Step 4: Run a visibility analysis in AI search engines

Once you’ve mapped your prompt clusters, the next step is to test if your brand appears in responses.

You can:

• Paste prompts manually into ChatGPT, Gemini, Perplexity, or Claude

• Use tools like LLMO Metrics to simulate neutral sessions and measure response visibility

⚠️ Important: If you test prompts from an account that has search history or previous chats, results may be biased. Clear your session or use an anonymous browser to avoid personalization effects.

Step 5: Prioritize prompts by value, volume and opportunity

Some prompts will be more strategic than others.

Here’s how to prioritize:

• High intent: Questions closer to conversion

• Frequency: Prompts that users ask often (manually or via tools)

• Opportunity: Where your competitors appear, but you don’t

Tools like AthenaHQ or LLMO Metrics can surface high-impact prompts based on your sector and keyword inputs.

Step 6: Create content that LLMs want to quote

Once you know which prompts matter — and where you’re missing — create content that LLMs can easily extract, understand and cite.

Tips to increase your chances of inclusion:

• Answer the query clearly and directly

• Use structured formatting and factual tone

• Add context, credibility, and unique value

• Use plain, readable language (clarity over cleverness)

In short:

You’re not just writing for Google anymore.
You’re writing for a machine that decides which brands are worth quoting.

Conclusion: Keyword research for the age of AI search

Keyword research hasn’t died. It has evolved.

In the AI era, we no longer optimize for keyword density or backlinks.
We optimize to become the most relevant answer.

That means:

• Starting from real user pain points

• Understanding the intent behind how people ask

• Mapping and testing prompts, not just keywords

• Creating high-authority, machine-readable content

Keyword research for GEO is about showing up in the answers — not just the rankings.

And if your brand isn’t showing up today, now is the time to fix it.