Most brands chasing AI search visibility are asking the wrong question.
They want to know how to rank inside ChatGPT, Perplexity, or Claude. That is not the right starting point. The better question is: what are your customers actually typing into those tools before they find anyone?
At MackyClyde SEO, this is where our GEO work starts. Not with guessing which keywords to stuff into a page. With understanding the specific prompts your market is already using to research, compare, and shortlist solutions.
This is harder than traditional keyword research. It is also more useful.
Traditional SEO trained businesses to think in two or three-word phrases. GEO requires you to think in full questions, layered problems, and comparison language. Someone is not typing “CRM software” into ChatGPT. They are typing something like, “What is the best CRM for a five-person sales team that needs automation but does not have budget for Salesforce?” That difference in phrasing changes everything about how content needs to be written.
Below are the three methods we use to uncover those prompts. Each one pulls from a different source, and together they give you a much clearer picture of how real people ask about your product category in AI environments.
Method 1: Mine Google Search Console Like a Prompt Researcher, Not an SEO
Search Console is not just a rankings dashboard. For GEO, it is one of the best sources of real customer phrasing you have access to.
Every query that brought someone to your site is something an actual person typed. Among those queries, the longer ones tend to look remarkably similar to how people phrase questions inside ChatGPT or Perplexity. The problem is that most teams glance at impressions and clicks, then close the tab.
Here is what we do differently.
We filter for longer queries, anything above five or six words, and treat them as intent signals rather than just search terms. A query like “how to improve AI visibility for a website” is not just a keyword. It is a user telling you exactly how they think about the problem, in their own words. That phrasing is research gold.
Once we have that list exported, we group them loosely by what kind of intent they reflect. Some are use-case driven, things like “how to get cited in ChatGPT answers” or “how to optimise content for Perplexity.” Some are comparison queries: “SEO vs GEO for AI search,” “best tools to track brand mentions in AI.” And some are audience-specific, where the same underlying question shows up differently by industry. “GEO for SaaS companies” reads very differently from “AI search strategy for ecommerce brands,” even if both are asking roughly the same thing.
After grouping, we expand each short query into a fuller prompt. Not to pad it out, but because the longer phrasing is closer to how the question will actually appear inside an AI tool.
For example:
Search query: geo for ecommerce Expanded prompt: What are the best GEO strategies for an ecommerce brand that wants more visibility in ChatGPT and AI search tools?
Search query: seo vs geo Expanded prompt: What is the difference between SEO and GEO, and which one should a business prioritise for AI-driven search visibility?
This is how we connect existing search data to GEO content planning. Most of the prompt opportunities are already sitting inside data a brand already owns. They just need to be read differently.
Method 2: Use AI Tools to Expand What You Already Know
Once you have seed prompts from Search Console, AI tools themselves can help you go deeper.
This sounds obvious, but most people do not use it this way. They use ChatGPT to write content. We use it to map the territory around a topic, specifically to find all the follow-up questions a user might ask before making a decision.
How we use Perplexity for this:
Start with a seed prompt that reflects a real customer concern. Something like: How can a company improve its visibility in ChatGPT?
Then pay attention to what comes next. What questions does Perplexity surface? What does it link to? What does it treat as related?
You might get follow-ups like:
- How does AI decide which sources to cite?
- Does website authority affect AI-generated recommendations?
- What makes one brand more visible in AI answers than another?
Each of those is a content opportunity. More importantly, each one tells you something about how a prospect thinks through this problem before they are ready to talk to anyone.
How we use ChatGPT to build prompt clusters:
We ask it directly. Prompt it to generate twenty follow-up questions around a GEO topic, then split those by intent: beginner questions, commercial questions, objection-led questions.
Starting from: How do I optimise my website for AI search?
You get questions like:
- Does structured content improve visibility in AI answers?
- Why do some websites get cited in ChatGPT while others do not?
- Can a small business compete in AI search without a large brand footprint?
- How do I measure whether GEO is actually working?
That last one matters a lot. It is the kind of question someone asks after they have already started caring about AI visibility but before they are confident enough to commit budget. That is a mid-funnel prompt, and it needs different content than an introductory “what is GEO” page does.
The prompts at different stages of a research journey map to different content types. When you know that mapping, you can build pages with a clear job to do instead of vague thought leadership that never quite lands. That is something we try to get right early in any GEO engagement.
Method 3: Use Reddit to Hear How Real People Actually Phrase Their Problems
Search Console gives you real queries. AI tools help you expand them. Reddit gives you something neither can: unfiltered language.
People on Reddit are not trying to impress anyone. They describe problems the way they think about them, not the way a marketer would frame them. That messiness is useful because AI prompts are often equally messy: casual, half-formed, specific in unexpected ways.
The best subreddits are not always the obvious ones. For GEO specifically, founder communities and small business subs often produce better prompt research than technical SEO communities, because users there ask outcome-based questions rather than jargon-heavy ones.
Look for posts like:
- “Why is my brand not showing up in AI tools?”
- “Anyone actually getting leads from ChatGPT traffic?”
- “Why does my content rank in Google but never get mentioned in AI?”
These are not keywords. They are the actual questions your potential clients are already asking in public. The phrasing matters too. “Why is my brand not showing up” is a different emotional entry point than “how to improve AI visibility,” even if the underlying topic is the same.
We refine those raw posts into cleaner prompts:
Reddit post: Anyone actually getting leads from ChatGPT traffic? Refined prompt: Can businesses generate qualified leads from ChatGPT and AI search traffic, and what makes it more likely?
Reddit post: Why is AI never mentioning my brand? Refined prompt: What are the reasons a brand may not appear in AI-generated answers, even if it has content and search visibility?
The original post is often just the surface, though. Comments reveal more nuance than any keyword tool will. We have seen threads produce prompts like:
- “How should content be structured to improve the chances of being referenced by AI tools?”
- “Why can a website rank well in Google but still fail to appear in ChatGPT answers?”
- “What GEO strategies work specifically for B2B companies?”
That third one is particularly telling. It shows someone who has already heard about GEO but does not trust that the general advice applies to their context. That is a content gap worth filling, and the kind of thing that rarely shows up in a keyword planner.
Turning Prompt Research Into Something You Can Actually Use
Running through these three methods once is useful. Running through them systematically, then organising what you find, is where it becomes a real strategic tool.
We build what we call a prompt library for each client, a document that captures not just the prompt itself, but where it came from, what intent it reflects, which stage of the buyer journey it sits in, and what content it should support or connect to.
This changes how content decisions get made. Instead of asking “what should we write about this month,” the question becomes “which prompts do we not yet have answers for, and which ones are closest to where someone decides?”
The difference matters. A lot of brands publish content that is technically correct and genuinely unhelpful because it does not map to how real people are thinking at the moment they are searching. GEO content needs to do a specific job at a specific point in someone’s research. Prompt research is how you figure out what those jobs actually are.
What We Have Learned Running This for Clients
A few things have surprised us doing this work.
The most commercially valuable prompts are rarely the most obvious ones. Generic searches like “best SEO agency” are competitive and vague. But a prompt like “which SEO agency is better for AI visibility and technical optimisation” reflects someone much further along in their thinking, and that specificity is exactly what good GEO content should target.
The gap between how brands describe their services and how prospects actually search for them is also usually wider than expected. Prompt research makes that gap visible. That alone tends to change how clients think about their content strategy, not just for AI search, but across the board.
The best starting point is simple. Pull your Search Console data this week. Export the long-tail queries. Look at them not as keywords but as questions your market is already asking, and ask yourself honestly whether you have a page that answers each one well.
Most of the time, the answer is no. Not because the content is bad, but because it was written around what the brand wanted to say, not around the specific questions people ask when they have not decided yet.
That is the gap prompt research helps close. It is also, if we are being honest, the most underused part of most GEO strategies we see.
