This guide explains how to create a structured set of prompts for tracking you AI visibility. You’ll learn how to define key topics, validate them, generate effective prompts.
The biggest challenge in GEO now is tracking prompts without knowing if they drive any traffic or if my website is being referenced.
Selecting the right prompts to track is one of the most crucial steps in Generative Engine Optimization and often the hardest.
Indeed, major AI models (ChatGPT, Claude, Gemini, etc.) currently provide no data on query volumes or on the types of queries users actually make.
This is a structural limitation for any GEO strategy.
Prompt identification therefore has to be done through proxies, because native data does not yet exist.SEO research is often an excellent starting point.
Step 1: Define Your Core Topics #
Core topics are the high-level themes that shape your monitoring strategy in Getmint.
They represent the main areas where you want your brand to gain visibility and be referenced across AI-generated content.
Examples of Core Topics
- Products or Services: project management software, online grocery delivery, skincare products
- Customer Use Cases or Needs: “tools for remote collaboration,” “where to order groceries in under an hour,” “face cream for sensitive skin”
- Industry Trends: AI in workplace productivity, sustainable packaging, growth of quick commerce
- Competitors or Alternatives: “Company vs Company,” “Instacart alternatives,” “Alternatives to Sephora”
Step 2 : Create your strategic mapping #
For each topic, document these 4 elements:
- List your use cases. Example for marketing software: “automate email sending,” “manage my customer and prospect database,” “automate content creation.”
- List your targets
- List your competitors
- List your transactional keywords: Select 3 to 7 transactional keywords — those that generate the most business. These may be your SEO keywords or the keywords that generate the most searches or conversions on Google Ads.
Step 3 : Create your prompts #
Prompts are the core of Getmint. Everything we analyze begins with the questions you define. This is not traditional SEO keyword tracking. Insights will take time to become statistically significant, and results may vary day-to-day because AI models are inherently dynamic.
Why Prompts Matter ? #
AI models don’t rely on keywords: they respond to full, natural questions.
For example, instead of searching “best running shoes”, users ask “What are the best running shoes for marathon training?”
Visibility in AI means appearing in these conversations, either as a recommended brand.
Prompts simulate the real questions your audience might ask.
GetMint runs these prompts across multiple AI models, providing insights on how the models respond and which brands or content are highlighted.
Start with your transactional prompts #
Method: Keywords + Keywords × Target
Create variations by combining your transactional keywords with your different target audiences.
Example:
- What is the best automated emailing software?
- What is the best automated emailing software for small marketing teams?
Expected outcome: 15–20 transactional prompts per topic
Use a “job to be done” approach #
The idea is not to start from a product, solution, or service, but from the problem and the action the user wants to perform.
Transactional approach: Marketing emailing software
Job-to-be-done approach: I’m a marketing director and I need to regularly send emails to all my customers — what should I do?
Method: Use cases × Targets
For each use case, formulate action-oriented questions.
Expected outcome: 8–10 Job-to-be-Done prompts per topic
Recommended total prompt volume: 20–40 prompts per topic
Other good proxies you can use #
If you lack information about your transactional keywords, or if you haven’t done any work on your SEO yet.
1. Listen and Ask Your Customers #
Start by paying close attention to the language your customers use when talking about your product or the challenges you help them solve.
- Analyze existing interactions: Review support tickets, chat transcripts, emails, or FAQs to spot recurring questions and common pain points.
- Engage your team: Speak with your sales and customer success teams to understand how prospects and clients describe their needs and frustrations.
- Leverage real conversations: The most effective prompts often originate directly from authentic customer interactions, capturing the exact words and concerns they express.
- Understand their journey: Ask how they discovered your product or service. This helps you identify which channels and touchpoints are most effective.
- Explore AI interactions: For customers coming through AI-driven channels, ask what queries or prompts they used when interacting with AI.
- Use their input to refine prompts: Their questions reveal exactly what information they’re seeking, allowing you to craft prompts that guide prospects more effectively through the buying process
2. Tap Into Public Questions #
AI models often learn from public conversations, so observing how people discuss topics online can reveal how users naturally phrase their intent.
- Scan relevant communities: Search Reddit, X (formerly Twitter), or product-specific forums related to your category.
- Focus on questions and threads: Pay attention to recurring questions, discussion threads, and problem-focused posts that relate to your space.
3. Get Insights from Reviews #
Customer reviews offer more than insights into satisfaction—they reveal the language users naturally associate with your brand and your competitors.
- Analyze multiple platforms: Check reviews on Trustpilot, G2, Appvizer, or other relevant sites for both your product and competing solutions.
- Focus on phrasing and comparisons: Take note of how users describe features, pain points, and alternatives
In conclusion: Start small. It’s better to identify five real prompts and iterate than to follow lots of prompts that don’t match your target audience’s needs. Then, expand your scope as you discover new questions and patterns in the AI’s responses.
Regularly run your most effective prompts. When you identify recurring patterns in the AI’s responses, use this information to refine your content.
Best Practices #
Always keep in mind that a visibility prompt is a question designed to track whether AI systems mention your brand in their responses.
Example of a visibility prompt:
“Which AI tools help brands improve their visibility on ChatGPT?” — this prompt can generate responses where your brand is cited.
Example of a prompt that is NOT a visibility prompt:
“What is visibility in AI?” — this asks a general informational question and does not track whether your brand is mentioned.
Think in Topics, Not Keywords #
Group related questions under broader strategic topics instead of tracking isolated keywords.
Each topic should represent a clear business theme tied to your brand, product, or market positioning.
Example:
❌ Instead of tracking variations of “AI marketing tool”
✅ Create a topic like “AI-Powered Marketing Platforms” with relevant prompts such as “What are the best AI tools for marketing teams?” or “Top AI platforms for lead generation.”
Focus on User Intent #
Write prompts that reflect natural language and real user intent.
Use full questions or conversational phrasing, as AI models interpret context-rich queries more accurately.
Example:
✅ “Which AI tools help brands improve their visibility in ChatGPT?”
❌ “AI visibility tools”
Prioritize Relevance Over Volume #
It’s better to track a smaller number of high-impact prompts than thousands of minor variations.
Each prompt should align with your strategic visibility goals and areas where you want your brand to be cited.
Example:
✅ “What are the best platforms to analyze brand visibility in AI?”
❌ “Brand visibility analysis tool”
