How Annapurna Selects Prompts for Tracking and Data Generation

The prompts you track matter.

In traditional search, businesses have spent years identifying the keywords their customers use on Google. AI search introduces a new layer. People are no longer searching only with short keywords. They are asking detailed questions, describing problems, comparing options, and asking AI platforms for recommendations.

That means tracking a large number of randomly generated prompts does not necessarily give a brand useful insight.

At Annapurna, we take a more deliberate approach. We combine human research, search data, customer intent, and AI-assisted analysis to identify prompts that are genuinely relevant to each business.

Our goal is simple: track the questions real customers are likely to ask and understand how visible your brand is within those conversations.

Here is how we approach it.

1. We Start With the Business, Not With Automation

Prompt research at Annapurna is not a fully automated process.

Before selecting prompts, we first try to understand the business itself: what it offers, who its customers are, where it operates, what problems it solves, and what alternatives customers may consider.

This context matters.

An automated system can generate hundreds of prompts within seconds, but volume alone does not make those prompts valuable. A technically relevant question may have little connection to how customers actually discover or evaluate a particular business.

We therefore use automation as a tool rather than allowing it to make every decision.

Our research process combines human judgment with AI and search data to build a prompt set that is specific to the brand and its market.

Businesses can also provide prompts they already know are important. If there are specific questions, services, products, or topics you want us to track, those can become part of the tracking strategy.

If you are not sure what to track, that is completely fine. Our team conducts the research and builds a set of at least 20 relevant prompts based on the business, its customers, search behavior, and market.

2. We Use Search Data to Understand What People Actually Look For

Relevance is important, but so is demand.

Two prompts can express almost the same idea while having very different levels of search interest. That is why traditional keyword and search-volume data remain useful when building an AI visibility strategy.

When evaluating potential prompts, we consider signals such as:

• Search volume: How frequently people search for the topic or phrase

• Keyword difficulty: How competitive the topic is

• Search intent: What the person is likely trying to accomplish

• Geographic relevance: Whether the query makes sense for the areas the business serves

• Business relevance: How closely the query connects to the company's actual products or services

Consider a family entertainment business.

A parent might ask:

"Where should I take my toddler for fun?"

Another might ask:

"Where can I take my 7-year-old for fun?"

Both questions are relevant, but they may have very different search demand.

Rather than treating every possible variation equally, we use available search data to understand which topics and phrases deserve greater attention.

The goal is not simply to find high-volume keywords. It is to find the intersection between search demand, customer intent, and business relevance.

 

3. We Track Different Stages of the Customer Journey

Customers do not always begin by searching for a company or even a specific service.

Sometimes they already know the brand. Sometimes they know exactly what service they need. Other times, they simply have a problem and are looking for guidance.

For that reason, Annapurna generally organizes prompts into three categories.

Type A: Brand-Aware Prompts

These prompts come from people who already know the brand and want additional information before making a decision.

For example:

"Tell me about Bounce House Fargo."

At this stage, the customer may be checking reputation, services, pricing, reviews, or whether the company is a good fit.

These prompts help us understand what AI platforms know and communicate about a brand when someone searches for it directly.

Type B: Solution-Seeking Prompts

These customers know what they need, but they may not know which company to choose.

For example:

"Plumber near me."

The user has already identified the solution. The question is which provider AI systems will recommend or mention.

These prompts are particularly important because they often represent strong commercial intent.

Type C: Problem-Aware Prompts

These users know something is wrong or that they have a need, but they may not yet know what product or service will solve it.

For example:

"Why is my car making a strange noise?"

The person may eventually need a mechanic, but they have not reached that conclusion yet.

Being visible during this stage can be valuable because the brand has an opportunity to become part of the customer's consideration process much earlier.

Together, these three categories give us a broader view of AI visibility across the customer journey, rather than measuring only direct brand searches or obvious commercial queries.

4. We Use AI to Refine Prompts, Not Simply Generate Them

AI is an important part of our research process, but we do not rely on it blindly.

Once we have identified promising topics using business research and search data, AI helps us explore how those topics might naturally appear in a conversation.

The distinction is important.

People often communicate differently with an AI assistant than they do with a traditional search engine. Someone might type:

"best plumber austin"

into a search engine, while asking an AI assistant:

"Who is a reliable plumber near me for a leaking pipe?"

The intent is similar, but the way the question is expressed is different.

We use AI to help identify these conversational variations and evaluate whether the prompts sound like questions real customers might reasonably ask.

Our team then reviews and refines the final selections.

This gives us a practical balance between human judgment, traditional search intelligence, and emerging AI search behavior.

5. We Finalize the Prompt Set and Begin Data Generation

Once the research is complete, we finalize the prompts that will be used for tracking.

From there, Annapurna begins the data generation process around those prompts.

Rather than treating the entire prompt set as one broad dataset, we work through the prompts individually. This allows us to build information that is closely aligned with the specific questions and topics where the brand wants greater visibility.

The result is a more structured approach to data generation.

Each prompt has a purpose. Each prompt represents a potential customer conversation. And the data we generate is designed to strengthen the information available about the brand within those relevant contexts.

Visibility tracking for all pompts.

Why We Take This Approach

AI visibility is still an emerging field, and it can be tempting to measure success by tracking as many prompts as possible.

We believe relevance is more valuable than volume.

Tracking hundreds of loosely related prompts may produce a larger dashboard, but it does not necessarily produce better business intelligence.

A smaller, carefully researched set of prompts can provide a much clearer picture of where a brand appears, where it is missing, and which areas deserve additional attention.

That is why Annapurna combines several signals rather than relying on a single automated system:

Business understanding + human research + search data + customer intent + AI-assisted analysis.

The objective is not to predict every question someone could possibly ask.

It is to identify the questions that matter most to the business and then build a strategy around improving visibility for those conversations.

As AI-powered discovery continues to evolve, we believe this combination of human judgment and data-driven research provides brands with a more practical and meaningful way to understand their presence across AI platforms.

If you already know the prompts that matter to your business, you can bring them to us. If you do not, our team will research them for you.

Either way, the goal remains the same: help your brand become visible when potential customers are asking the questions that matter.