Setting Up Prompts and Topics: The Basics
After this lesson, you'll have a structured tracking system in AirOps that shows exactly where your brand appears in AI search and where you're invisible. Dillon walks you through setting up topics and prompts so you can replace keyword lists and gut instinct with real data on whether answer engines are recommending you.
What You'll Learn
By the end of this lesson, you'll know how to organize the questions buyers ask AI into meaningful topics and build prompts that reflect real commercial intent. Dillon walks through the Intelligence phase of the CITED framework and shows you how to set up your tracking foundation in AirOps.
From keywords to prompts
Most AI prompts run 10 words or longer. 60% are 10+ words, and the single largest bucket is 21+ words.
People aren't typing fragments into AI. They're describing problems and asking for recommendations.
The unit of tracking shifts from keywords to buyer questions. Keywords still help you find topics, but they don't become your prompt list directly.
Topics and prompts: the mental model
- A topic is a collection of related prompts that share a coherent area of intent. Topic-level metrics (mention rate, citation rate, share of voice) aggregate from the prompts inside.
- A prompt is a specific question sent to an answer engine.
- Keep topics focused. Mixing unrelated prompts into one topic turns your metrics into noise and strips out their diagnostic value.
What makes a good topic
Prompts in the same topic should share the same competitive set and buying criteria. They should also point toward a single content strategy.
When competitors or criteria shift significantly, split into separate topics.
Dillon uses Webflow as an example: "website builders" and "enterprise CMS platforms" are two different topics because the competitor set changes entirely. Squarespace and Wix show up in one. Contentful and Sitecore show up in the other.
Five sources for mapping your topics
- Capabilities and product lines. What your product does, broken into distinct functional areas.
- Buyer roles and teams. Who is making or influencing the purchase decision.
- Buying criteria. The specific requirements buyers evaluate (speed, integrations, ease of use).
- Hard limits. Non-negotiable constraints like budget, compliance requirements, or tech stack compatibility.
- Strategic moments. Time-bound opportunities such as product launches or campaigns.
Coverage topics vs. depth topics
Coverage topics are broad. They monitor whether you're part of the conversation for an entire category. Aim for 10-15 coverage topics. These tell you where you show up and where competitors dominate.
Depth topics are narrow. They're designed to win specific opportunities with a handful of focused prompts. The question they answer: "Are we the default recommendation for this use case?"
The best teams build both. Start with coverage topics to establish a baseline, then add depth topics for the 2-5 areas driving the most revenue.
How to construct prompts: four dimensions
Every commercial prompt starts with a base (your category term) and adds dimensions to reflect real buyer intent:
- Buyer. Who's deciding. Role, team type, industry.
- Outcome. What they need. Capabilities, features, specific results.
- Boundary. What limits the decision. Budget, compliance, tech stack.
- Benchmark. What they're weighing you against.
For coverage topics, Dillon recommends base + one dimension. For depth topics, stack two or three dimensions to match the specificity of real buyer queries.
How depth feeds coverage
Answer engines break broad questions into sub-queries behind the scenes. When you win on specific depth queries, you're more likely to show up in broad coverage answers too. Depth work compounds upward.
Setting it up in AirOps
Dillon walks through the AirOps platform to show how to navigate your prompts and topics and how to add new ones. This is where the concepts from the lesson become a working setup you can build on.
Key takeaways
- 21+ words is the new normalThe single largest category of AI prompts is 21 words or more. Buyers aren't searching in fragments. They're describing full scenarios, which means your tracking system needs to match that specificity.
- Your keyword list is an input, not a prompt listKeywords tell you what people care about, but buyers don't talk to answer engines in two-word phrases. Use your keyword research to identify topics, then translate those topics into the full questions buyers actually ask.
- Mixed topics produce meaningless metricsTopic-level metrics aggregate from every prompt inside. Combining unrelated prompts under one topic turns your mention rate and citation rate into noise. Keep each topic to a single competitive set and buying criteria.
- Base + dimensions = a prompt that forces a recommendationEvery commercial prompt starts with a category term, then adds buyer, outcome, boundary, or benchmark dimensions. Coverage topics use one dimension. Depth topics stack two or three. The more specific the prompt, the more it mirrors how real buyers ask for recommendations.
- Winning narrow queries compounds into broad visibilityAnswer engines decompose broad questions into sub-queries before assembling a response. When you win on specific depth prompts, those wins feed into your broad coverage answers through sub-query compounding.
FAQs
A topic is a group of related prompts that share the same competitive set and buying criteria. A prompt is a specific question you send to an answer engine to test whether your brand appears in the response. Topic-level metrics like mention rate and citation rate aggregate from the prompts inside, so keeping topics focused ensures your data stays actionable.