3/7 Mastering Prompt Research for Effective Brand Monitoring
In this lesson (5-min read + video): how to build the prompt set everything else depends on — what makes a good prompt, where to find real ones, and how many you need.
Video link: https://www.loom.com/share/4cc9e24bcbf546068e08049f3dfe749a
Key concepts
Prompt — a question a customer might ask an AI engine to find products or services like yours (e.g. "best CRM for a 20-person SaaS team"). Prompts are the new keywords.
Non-branded prompt — a question that doesn't name your brand. Branded questions bias the results, so you monitor unbranded ones to see whether you show up on your own merits.
Bottom-of-funnel intent — questions from someone close to choosing ("best running shoe for flat feet"), not top-of-funnel definitions ("what is a running shoe?"), which rarely return brands.
Intent volume — Otterly's estimate of how much real demand sits behind a prompt, based on Google search volume (there's no official "prompt volume" yet).
Quick reference
The golden rule. Non-branded, middle-to-bottom-of-funnel questions only. Skip brand names and skip "what is X?" — both waste monitoring.
How many. Aim for ~100 prompts per country, per language, per product category; 50 is the floor, never fewer. Two categories × two countries × two languages ≈ 400 prompts.
Where real prompts come from. AI Prompt Research (from your keywords, URL, or a competitor's), Query Fan-Out for variations, Google/Bing Search Console, customer surveys, and your sales and support conversations.
Organize as you go. Tag every prompt by product, funnel stage, country, and language so you can segment reports later. Keep the master list in Otterly or a spreadsheet.
Quality bar. A good prompt is a real question, ~6–12 words, implies buying or evaluation intent, and realistically returns brands when asked.
⚠️ Deleting a prompt can't be undone — you lose its history too. Double-check before removing any.
Your checklist for this lesson
➡️ Next: Lesson 4 — Read your brand report.